Body Wasting Among Tuberculosis Patients in Urban Uganda

Home , Wasting

BODY WASTING AMONG TUBERCULOSIS PATIENTS IN URBAN UGANDA

EZEKIEL MUPERE MBChB, M.MED (Paediatrics), M.S

Submitted in partial fulfillment of the requirements

For the degree of Doctor of Philosophy

Dissertation Advisor: Dr. Daniel Tisch PhD, MPH

Department of Epidemiology and Biostatistics

CASE WESTERN RESERVE UNIVERSITY

May, 2010

DEDICATION

My wife Harriet Mupere Babikako for your cheerful love, prayers,

and persuasive encouragement

My children Patience Esther Mupere, Elizabeth Peace Mupere, Ednah Precious Mupere,

and those I prayerfully hope to be born, for being a strong encouragement. Desire to

embrace this academic path

My elder brother Daniel Bawunha for establishing the foundation of my academic path

and for your words of wisdom

And to,

The memory of my late mother Mary Dorothy Wandawa for the exemplary parental love

and vision, and for introducing me to the objective of learning a, e, i, . . .., and to the objective of learning how to write a love letter; now accomplished. I’m forever indebted.

TABLE OF CONTENTS

LIST OF TABLES ...... vi

ACKNOWLEDGEMENT ...... xx

Abstract ...... xxiii

CHAPTER 1 ...... 1

INTRODUCTION ...... 1

CHAPTER 2 ...... 15

BACKGROUND ...... 15

CHAPTER 3 ...... 32

STUDY METHODS ...... 32

CHAPTER 4 ...... 46

BODY COMPOSITION MEASURED WITH BIOELECTRICAL IMPEDANCE

ANALYSIS AND ANTHROPOMETRY AMONG HIV POSITIVE AND HIV

NEGATIVE ADULTS WITH OR WITHOUT TUBERCULOSIS IN KAMPALA,

UGANDA...... 46

CHAPTER 5 ...... 84

INDICATORS OF DIETARY ADEQUACY AMONG HIV POSITIVE AND HIV

NEGATIVE ADULTS WITH OR WITHOUT TUBERCULOSIS IN URBAN

KAMPALA, UGANDA ...... 84

CHAPTER 6 ...... 126

iii

PREDICTORS OF FAT MASS AND LEAN TISSUE AMONG HIV POSITIVE AND

HIV NEGATIVE ADULTS WITH OR WITHOUT TUBERCULOSIS IN URBAN

KAMPALA, UGANDA ...... 126

CHAPTER 7 ...... 169

BODY WASTING AND DIETARY INTAKE AMONG HIV POSITIVE AND HIV

NEGATIVE ADULTS WITH OR WITHOUT TUBERCULOSIS IN URBAN

UGANDA, KAMPALA ...... 169

CHAPTER 8 ...... 202

CORRELATES OF DIETARY INTAKE AMONG HIV POSITIVE AND HIV

NEGATIVE ADULTS WITH OR WITHOUT TUBERCULOSIS IN URBAN

KAMPALA, UGANDA ...... 202

CHAPTER 9 ...... 233

IMPACT OF BODY WASTING ON SURVIVAL AMONG ADULT PATIENTS

WITH PULMONARY TUBERCULOSIS IN URBAN KAMPALA, UGANDA ..... 233

CHAPTER 10 ...... 292

LONGITUDINAL CHANGES IN BODY COMPOSITION AMONG HIV POSITIVE

AND HIV NEGATIVE ADULT PATIENTS WITH PULMONARY

TUBERCULOSIS IN URBAN KAMPALA, UGANDA ...... 292

CHAPTER 11 ...... 347

SUMMARY OF RESULTS AND GENERAL CONCLUSIONS ...... 347

CHAPTER 12 ...... 365

EXTRA TABLES AND FIGURES FOR CHAPTER FOUR ...... 365

CHAPTER 13 ...... 377

EXTRA TABLES AND FIGURES FOR CHAPTER NINE ...... 377

CHAPTER 14 ...... 414

EXTRA TABLES AND FIGURES FOR CHAPTER TEN ...... 414

BIBLIOGRAPHY ...... 441

LIST OF TABLES

Table 4:1 Characteristics of study population with tuberculosis ...... 65

Table 4:2 Characteristics of study population without tuberculosis ...... 67

Table 4:3 Comparison fat-free mass and fat mass as measured by BIA and by equations

that involved anthropometry measurements for all participants (n=131) ...... 69

Table 4:4 Comparison fat-free mass and fat mass measured by BIA and by equation with

waist circumference ...... 70

Table 4:5 Comparison fat-free mass and fat mass measured by BIA and by equation with

sum of 4 skinfolds ...... 72

Table 4:6 Comparison fat-free mass and fat mass measured by BIA and by equation with

BMI ...... 74

Table 4:7 Comparison fat-free mass and fat mass measured by BIA and by equation with

MUAC...... 76

Table 5:1 Characteristics of adult participants with tuberculosis in Kampala, Uganda . 104

Table 5:2 Characteristics of adult participants without tuberculosis in Kampala, Uganda

...... 105

Table 5:3 Food groups and food items from 24-hour dietary intake recall among HIV positive and HIV negative adults in Kampala, Uganda (n=131) ...... 106

Table 5:4 The 24-hour dietary intake recall, recommended daily allowance, frequency of inadequate intake of nutrients among HIV positive and HIV negative adults in Kampala,

Uganda (n=131) ...... 108

Table 5:5 Percent of inadequate 24-hour dietary recall intake among HIV positive and

HIV negative adults with tuberculosis in Kampala, Uganda ...... 110

Table 5:6 Percent of inadequate 24-hour dietary recall intake among HIV positive and

HIV negative adults without tuberculosis in Kampala, Uganda ...... 112

Table 5:7 Spearman’s correlations between Nutrient Adequacy Ratio (NAR) of nutrients and Food Variety Score or Dietary Diversity Score among HIV positive and HIV negative adults with or without tuberculosis in Kampala, Uganda (n=131)...... 114

Table 5:8 Mean MAR scores for different levels of Food Variety Score (FVS) and

Dietary Diversity Score (DDS) among HIV positive and HIV negative adults in Kampala,

Uganda (n=131) ...... 116

Table 5:9 Estimated Mean Adequacy Ratio scores for different levels of Food Variety

Score (FVS) and Dietary Diversity Score (DDS) from Linear Regression Model among

HIV positive and HIV negative adults in Kampala, Uganda (n=131) ...... 118

Table 6:1 Characteristics of the study population (n=131) ...... 145

Table 6:2 Nutrient intake characteristics of the study population (n=131) ...... 148

Table 6:3 Spearman’s correlations between energy or protein intake and body mass index, fat or fat-free mass (n=131) ...... 149

Table 6:4 Predictors of fat-free mass in univariate models among HIV positive and HIV negative adults with or without tuberculosis in Kampala, Uganda (n=131)...... 150

Table 6:5 Predictors of fat-free mass in univariate models among HIV positive and HIV negative adults with or without tuberculosis in Kampala, Uganda stratified by sex

(n=131) ...... 152

vii

Table 6:6 Predictors of fat-free mass in multivariable models among HIV positive and

HIV negative adults with or without tuberculosis in urban Kampala, Uganda (n=131) 154

Table 6:7 Predictors of fat-free mass in multivariable models among HIV positive and

HIV negative adults with or without tuberculosis in Kampala, Uganda stratified according sex (n=131) ...... 155

Table 6:8 Predictors of fat mass in univariate models among HIV positive and HIV negative adults with or without tuberculosis in Kampala, Uganda (n=131)...... 156

Table 6:9 Predictors of fat mass in univariate models among HIV positive and HIV negative adults with or without tuberculosis in Kampala, Uganda stratified according to sex (n=131) ...... 157

Table 6:10 Predictors of fat mass in multivariable models among HIV positive and HIV negative adults with or without tuberculosis in Kampala, Uganda (n=131)...... 159

Table 6:11 Predictors of fat mass in multivariable models among HIV positive and HIV negative adults with or without tuberculosis in Kampala, Uganda (n=131)...... 161

Table 6:12 Predictors of body mass index in univariate models among HIV positive and

HIV negative adults with or without tuberculosis in Uganda (n=131) ...... 163

Table 6:13 Predictors of body mass index in univariate models among HIV positive and

HIV negative adults with or without tuberculosis in Kampala, Uganda stratified according sex (n=131) ...... 165

Table 6:14 Predictors of body mass index in multivariable models among HIV positive and HIV negative adults with or without tuberculosis in urban Uganda (n=131) ...... 167

Table 6:15 Predictors of body mass index in multivariable models among HIV positive and HIV negative adults with or without tuberculosis in Kampala, Uganda (n=131) .... 168

viii

Table 7:1 Select characteristics among HIV positive and HIV negative adults with tuberculosis ...... 186

Table 7:2 Select characteristics among HIV positive and HIV negative adults without tuberculosis ...... 188

Table 7:3 Dietary intake of 24-hour recall among HIV positive and HIV negative adults with tuberculosis ...... 190

Table 7:4 Dietary intake of 24-hour recall among HIV positive and HIV negative adults without tuberculosis ...... 192

Table 7:5 Select characteristics among patients with/without wasting (n=63)...... 194

Table 7:6 Select characteristics among patients with/without severity of clinical tuberculosis ...... 196

Table 7:7 Dietary intake of 24-hour recall among tuberculosis patients with/without body wasting (n=63) ...... 198

Table 7:8 Dietary intake of 24-hour recall among tuberculosis patients with/without severe clinical TBscore (>5) ...... 200

Table 8:1 Characteristics of the study population (n=131) ...... 215

Table 8:2 Correlates of dietary energy intake among women and men ...... 218

Table 8:3 Correlates of dietary protein intake among women and men ...... 221

Table 8:4 Proportions of adult individuals with inadequate dietary intakes of energy, protein, and micronutrients in relation to socio-demographic, HIV, and tuberculosis variables in Kampala, Uganda ...... 223

Table 8:5 Proportions of adult individuals with inadequate dietary intakes of energy,

protein, and micronutrients in relation to socio-demographic, HIV, and tuberculosis

variables in Kampala, Uganda ...... 226

Table 8:6 Correlates of inadequate dietary intakes of key nutrients from multivariate

models among adult individuals in Kampala, Uganda ...... 229

Table 9:1 Definitions of low and normal fat and fat-free mass index values for corresponding body mass index in adults ...... 243

Table 9:2 Mean body mass, fat and fat-free mass indexes; spearman correlations between body mass index and fat or fat-free mass indexes among adult women and men in urban

Uganda ...... 264

Table 9:3 Concordance between low body mass index and low fat-free or fat mass

indexes corresponding to body mass index in assessing wasting among adults in urban

Uganda ...... 265

Table 9:4 Baseline characteristics of pulmonary tuberculosis patients with normal versus

malnutrition ...... 267

Table 9:5 Stratified Analysis of mortality among tuberculosis with normal (≥18.5) and

low (<18.5) body mass index (BMI kg/m2) according to key variables [deaths/number at risk (%)] ...... 270

Table 9:6 Stratified Analysis of mortality among tuberculosis with normal and low fat- free mass index (FFMI kg/m2) according to key variables [deaths/number at risk (%)] 272

Table 9:7 Univariate Analysis of factors associated with survival ...... 274

Table 9:8 Relative hazards [HR, 95% confidence intervals (CIs)] for death among

tuberculosis patients with normal compared with patients having low body mass index

(BMI) ...... 277

Table 9:9 Relative hazards [HR, 95% confidence intervals (CIs)] for death among

tuberculosis patients with normal compared with patients having low body mass index

(BMI) stratified according to sex status ...... 279

Table 9:10 Relative hazards [HR, 95% confidence intervals (CIs)] for death among

tuberculosis patients with normal compared with patients having low body mass index

(BMI) ...... 281

Table 9:11 Relative hazards [HR, 95% confidence intervals (CIs)] for death among

tuberculosis patients with normal compared with patients having low body mass index

(BMI) stratified according to HIV status ...... 283

Table 9:12 Relative hazards [HR, 95% confidence intervals (CIs)] for death among

tuberculosis patients with normal compared with low fat-free mass index (FFMI) ...... 285

Table 9:13 Relative hazards [HR, 95% confidence intervals (CIs)] for death among

tuberculosis patients with normal compared with low fat-free mass index (FFMI)

stratified according to sex status ...... 287

Table 10:1 Definitions of low and normal fat and fat-free mass index values for corresponding body mass index in adults ...... 302

Table 10:2 Baseline characteristics of the study population with/without baseline wasting

...... 328

Table 10:3 Unadjusted rate of change for fat-free mass (FFMI), fat mass (FMI), and body mass index (BMI) among pulmonary patients with reduced FFMI, FMI, and BMI in

Kampala, Uganda ...... 330

Table 10:4 Adjusted rate of change for fat-free mass (FFMI), fat mass (FMI), and body

mass index (BMI) among tuberculosis patients presenting with reduced FFMI, FMI, and

BMI in Kampala, Uganda ...... 333

Table 10:5 Adjusted rate of change for fat-free mass (FFMI), fat mass (FMI), and body

mass index (BMI) among tuberculosis patients presenting with reduced FFMI, FMI, and

BMI stratified according to gender in Kampala, Uganda ...... 336

Table 10:6 Adjusted rate of change for fat-free mass (FFMI), fat mass (FMI), and body

mass index (BMI) among tuberculosis patients presenting with reduced FFMI, FMI, and

BMI in Kampala, Uganda ...... 339

Table 10:7 Adjusted rate of change for fat-free mass (FFMI), fat mass (FMI), and body

mass index (BMI) among tuberculosis patients presenting with reduced FFMI, FMI, and

BMI stratified according to HIV in Kampala, Uganda ...... 342

Table 12:12:1 Mean intra-supervisor, mean intra-observer, and mean inter-observer

technical error of measurement prior to data collection ...... 366

Table 12:12:2 Coefficient of reliability prior to data collection ...... 367

Table 12:12:3 Guide for comparison of inter-observer error (Frisancho A.R 1990) ...... 368

Table 13:1 Comparison of key baseline variables across phase II prednisolone study,

household contact study, and Kawempe community health study ...... 378

Table 13:2 Assessing the proportional hazards assumption with a statistical test:

Correlations between ranked failure time and Schoenfeld residuals ...... 380

xii

Table 13:3 Baseline characteristics of pulmonary tuberculosis patients between HIV

negative and HIV positive ...... 381

Table 13:4 Multivariable Relative hazards [HR, 95% confidence intervals (CIs)] for death

among tuberculosis patients with normal compared with patients having low body mass

index (BMI) ...... 384

Table 13:5 Multivariable Relative hazards [HR, 95% confidence intervals (CIs)] for death

among tuberculosis patients with normal compared with patients having low body mass

index (BMI) ...... 386

Table 13:6 Relative hazards [HR, 95% confidence intervals (CIs)] for death among

tuberculosis patients with normal compared with patients having low body mass index

(BMI) ...... 388

Table 13:7 Relative hazards [HR, 95% confidence intervals (CIs)] for death among

tuberculosis patients with normal compared with patients having low body mass index

(BMI) stratified according to Age group ...... 390

Table 13:8 Relative hazards [HR, 95% confidence intervals (CIs)] for death among HIV

positive tuberculosis patients compared with HIV negative patients ...... 392

Table 13:9 Relative hazards [HR, 95% confidence intervals (CIs)] for death among HIV

positive tuberculosis patients compared with HIV negative patients stratified according to

BMI categories ...... 394

Table 13:10 Multivariable Relative hazards [HR, 95% confidence intervals (CIs)] for

death among tuberculosis patients with normal compared with patients having low fat- free mass index (FFMI) ...... 396

xiii

Table 13:11 Relative hazards [HR, 95% confidence intervals (CIs)] for death among

tuberculosis patients with normal compared with low fat-free mass index (FFMI) ...... 398

Table 13:12 Relative hazards [HR, 95% confidence intervals (CIs)] for death among

tuberculosis patients with normal compared with low fat-free mass index (FFMI)

stratified according to HIV status ...... 400

Table 13:13 Relative hazards [HR, 95% confidence intervals (CIs)] for death among

tuberculosis patients with normal compared with low fat-free mass index (FFMI) ...... 402

Table 13:14 Relative hazards [HR, 95% confidence intervals (CIs)] for death among

tuberculosis patients with normal compared with low fat-free mass index (FFMI)

stratified according to age group...... 404

Table 13:15 Relative hazards [HR, 95% confidence intervals (CIs)] for death among HIV

positive tuberculosis patients compared with HIV negative patients ...... 406

Table 13:16 Relative hazards [HR, 95% confidence intervals (CIs)] for death among HIV

positive tuberculosis patients compared with HIV negative patients stratified according to

fat-free mass index (FFMI) categories ...... 408

Table 14:1 Spearman raw correlation matrix for fat-free mass, fat mass, and body mass index over different visit time points ...... 414

Table 14:2 Regression estimates for effects of baseline wasting, lag FFMI measure, and

baseline characteristics on probability of missing ...... 416

Table 14:3 Regression estimates for effects of baseline wasting, lag FMI measure, and

baseline characteristics on probability of missing ...... 418

Table 14:4 Assessing contributions of polynomials, random intercepts, and random

slopes in mixed models ...... 422

xiv

Table 14:5 Assessing Piecewise and polynomial models; random intercepts, and random slopes in mixed models ...... 424

LIST OF FIGURES

Figure 1:1 Working tuberculosis – HIV – Malnutrition Model ...... 6

Figure 3:1 Retrospective cohort study flow ...... 41

Figure 3:2 Cross-sectional study flow diagram ...... 42

Figure 5:1 Sensitivity (sens) and specificity (spec) for different cut-off points of Food

Variety Score (FVS) among women: Mean Adequacy Ratio (MAR) changing from 60 to

75...... 122

Figure 5:2 Sensitivity (sens) and specificity (spec) for different cut-off points of Food

Variety Score (FVS) among men: Mean Adequacy Ratio (MAR) changing from 60 to 75

...... 123

Figure 5:3 Sensitivity (sens) and specificity (spec) % for different cut-off points of Diet

Diversity Score (DDS) among women: Mean Adequacy Ratio (MAR) changing from 60 to 75 ...... 124

Figure 5:4 Sensitivity (sens) and specificity (spec) % for different cut-off points of Diet

Diversity Score (DDS) among men: Mean Adequacy Ratio (MAR) changing from 60 to

75...... 125

Figure 9:1 Survival distribution among adult men presenting with wasted body mass

(BMI <18.5 kg/m2) compared to men with normal body mass in urban Kampala, Uganda

...... 289

Figure 9:2 Survival distribution among adult women with baseline fat-free mass wasting

(FFMI <16.7 kg/m2 for men, <14.6 kg/m2 for women) compared to women with normal fat-free mass in urban Uganda ...... 290

xvi

Figure 9:3 Survival distribution Men compared with Women among HIV positive tuberculosis patients in urban Uganda ...... 291

Figure 10:1 Overall Study flow diagram ...... 345

Figure 10:2 Study flow diagram for BIA data ...... 346

Figure 11:1 Overall conclusions: Tuberculosis – HIV – malnutrition model ...... 359

Figure 12:1 Fat-free mass measured by BIA compared with fat-free mass measured by equation involving waist circumference ...... 369

Figure 12:2 Fat mass measured by BIA compared with fat mass measured by equation involving waist circumference ...... 370

Figure 12:3 Fat-free mass measured by BIA compared with fat-free mass measured by

Durnin & Womersely equations involving sum of 4 skinfold thickness ...... 371

Figure 12:4 Fat mass measured by BIA compared with fat mass measured by Durnin &

Womersely equations that involves sum of 4 skinfold thickness ...... 372

Figure 12:5 Fat-free mass measured by BIA compared with fat-free mass measured by equation involving BMI ...... 373

Figure 12:6 Fat mass measured by BIA compared with fat mass measured by equation involving BMI ...... 374

Figure 12:7 Fat-free mass measured by BIA compared with fat-free mass measured by equation involving MUAC ...... 375

Figure 12:8 Fat mass measured by BIA compared with fat mass measured by equation involving MUAC ...... 376

Figure 13:1 Survival distribution among women with low baseline body mass index

(BMI) (<18.5 kg/m2) compared to women with normal BMI in Uganda ...... 410

xvii

Figure 13:2 Survival distribution among men with low (<16.7 kg/m2 for men, <14.6 kg/m2 for women) baseline fat-free mass index (FFMI) compared to men with normal

FFMI in Uganda ...... 411

Figure 13:3 Survival distribution among Men compared to Women tuberculosis patients in urban Uganda ...... 412

Figure 13:4 Survival distribution among Men compared with Women among HIV negative tuberculosis patients in Uganda ...... 413

Figure 14:1 Individual profiles for fat-free mass over time ...... 426

Figure 14:2 Individual profiles for fat mass index over time ...... 427

Figure 14:3 Individual profiles for BMI over time ...... 428

Figure 14:4 Box plots for fat-free mass index over time ...... 429

Figure 14:5 Box plots for fat mass index over time...... 430

Figure 14:6 Box plots for BMI over time ...... 431

Figure 14:7 Mean profile for FFMI among patients with compared to patients without baseline wasting ...... 432

Figure 14:8 Mean profile for FFMI among men compared to mean profile among women

...... 433

Figure 14:9 Mean profile for FFMI among HIV negative compared to mean profile among HIV positive patients ...... 434

Figure 14:10 Mean profile for FMI among patients with compared to patients without wasting ...... 435

Figure 14:11 Mean profile for FMI among men compared to mean profile among women

...... 436

xviii

Figure 14:12 Mean profile for FMI among HIV negative compared to mean profile among HIV positive patients ...... 437

Figure 14:13 Mean profile for BMI among patients with compared to patients without wasting ...... 438

Figure 14:14 Mean profiles for BMI among men compared to mean profile among women ...... 439

Figure 14:15 Mean profile for BMI among HIV negative compared to mean profile among HIV positive patients ...... 440

xix

ACKNOWLEDGEMENT

Several persons contributed in different ways to the successful completion of this dissertation – to them I am profoundly indebted. I wish to express my deep gratitude to:

The Fogarty AIDS International Training and Research Program (AITRP) grant number

TW000011 at Case Western Reserve University (CWRU) for funding my training and accomplishment of this dissertation under directorship of Prof. Christopher Whalen. I am grateful to the administrative staff of the AITRP grant: Alice Cantini and Grace Sivlar at

CWRU, Robert Kakaire and Xiaoping at University of Georgia for the all the administrative support.

The dissertation committee members: Dr. Daniel J. Tisch, Dr. Mark Schluchter, Dr.

Isabel Parraga, Dr. Catherine Stein, and Dr. Christopher Whalen for the guidance and constructive criticisms that enabled me design, analyze, and write-up this dissertation.

Specifically, Dr. Tisch for not only accepting to chair the committee at such a time when

Dr. Whalen had to leave for Georgia but also for the moral support and guidance in drafting the manuscripts; Dr. Schluchter for the design, statistical skills and knowledge that shaped the dissertation; Dr. Parraga for the design and technical expertise in nutrition that has made me envision nutrition epidemiology as a research career; Dr. Stein for accepting at to be a member on the committee at short notice and for aggressively reviewing all the analysis and write style of the manuscripts in this dissertation; and Dr.

Whalen for not only being my chair and research advisor before his transfer to Georgia

but for also introducing me to critical thinking and tenacity in research. Dr. Whalen, I

will remain forever indebted.

Dr. Harriet Mayanja-Kiiza, Professor at the College of Health Sciences Makerere

University Kampala, Uganda, for the mentorship and for the constructive suggestions and criticisms during the design and write-up of this dissertation.

The study staff: Cissy Nambejja, Hassard Sempeera, Ruth Balaba, and Isaac Nsereko for the support in primary data collection which would have been impossible without their input.

All the staff of the Uganda-Case Western Reserve University Research Collaboration and

all the staff of the Tuberculosis Research Unit (TBRU) at Case Western Reserve

University for collecting and maintaining clean data that enabled me accomplish the

survival and longitudinal aims in this dissertation. Special thanks to LaShaunda Malone

and Allan Chiunda who put together the datasets that were used for analysis; to Denise

Jonhson who performed the data management for the primary cross-sectional data

collection; to Dr. Henry W. Boom, the PI of TBRU, for the constructive criticisms and

for granting me permission to use the data; and to the grant N01-AI95383 and

HHSN266200700022C/ N01-AI70022 from the NIAID that provided funding to TBRU.

xxi

The fellow Fogarty trainees for the period January 2006 to May 2010, for the moral

support, constructive criticisms and suggestions in the development and write-up of this

dissertation.

My wife Harriet Mupere Babikako, for standing with me during the good and difficult times in my academic career.

My elder brother, Daniel Bawunha for being a good source of encouragement and

wisdom in my academic pursuit.

xxii

Body Wasting among Tuberculosis Patients in Urban Uganda

Abstract

EZEKIEL MUPERE

Background Although body wasting is a cardinal feature of tuberculosis, its etiology and

management is poorly understood; and its assessment is overlooked in research and in

clinical practice.

Objective We established whether body wasting modifies survival and body composition changes during and after tuberculosis treatment; whether HIV modifies dietary intake

among tuberculosis patients; whether dietary intake differs by wasting and severity of

disease status; and whether dietary intake influences body composition.

Methods Retrospective cohort and cross-sectional designs were employed. Height-

normalized body mass (BMI), fat-free mass (FFMI), and fat mass (FMI) indices and 24-

hour dietary intake recall were measured.

Results Body wasting was associated with reduced survival and the effect differed by

gender. FFMI was found to be a predictor of survival among women whereas BMI was

xxiii

among men. Wasting was associated with substantial linear increase in FFMI, FMI, and

BMI during the first three months but the rate of increase differed by gender and not HIV status. Changes in body composition among men were affected by initial FFMI and BMI, whereas among women by FMI. There were minimal changes in body composition after month 3 and during the one year period after month 12 regardless of the initial body composition, gender, and HIV status. Dietary intake in the study population was monotonous, rich in carbohydrates and deficient in nutrients. Dietary intake at the time of diagnosis was influenced by severity of tuberculosis disease, but not HIV status and in the absence of tuberculosis was influenced by gender. Prediction of body composition by energy and protein intake differed by gender. Energy intake was an important predictor of body composition among women whereas appetite was among men.

Conclusion Results provide theoretical framework to provide targeted nutritional intervention to patients presenting with wasting and patients of female gender. National programs should integrate nutritional health education in the management of tuberculosis.

Nutritional assessment should involve establishment of body composition to identify patients that may be at risk of poor survival. Further evaluation is needed to understand changes in dietary intake overtime and its impact on body composition.

Key words: Tuberculosis; body wasting; survival; body composition; dietary intake

Word count: 343

xxiv

CHAPTER 1

INTRODUCTION

Tuberculosis is the most relevant infectious disease worldwide, and its etiological agent,

Mycobacterium tuberculosis, infects one-third of the world population (World Health

Organization. Global tuberculosis control. Epidemiology, strategy, financing. WHO

2009). Globally, 9.2 million new cases and 1.7 million tuberculosis-associated deaths occurred in 2006 (Vitoria et al. 2009). In sub-Saharan Africa, where the prevalence and

incidence of tuberculosis are among the highest in the world, 79% adults with

tuberculosis are also infected with human immunodeficiency virus (HIV) (Lawn and

Churchyard 2009; World Health Organization 2009), resulting in elevated mortality.

Tuberculosis is frequently found at autopsy of acquired immunodeficiency syndrome

(AIDS) patients (Lucas et al. 1994; Lucas et al. 1993), particularly cachectic patients

suggesting that tuberculosis exacerbates the wasting process of HIV-infected people in

Africa.

Tuberculosis and HIV are both independently associated with body wasting (Scalcini et al. 1991; Suttmann et al. 1995; Harries et al. 1988; Kennedy et al. 1996). In tuberculosis-

HIV co-infection, there may be additional metabolic, physical, and nutritional burden, resulting in potential further increase in energy expenditure, malabsorption, and

micronutrient deficiency. Thus, the suggested tuberculosis or co-infection malnutrition

model (Figure 1:1). In this model because of differences in metabolism, gender differences in body composition at presentation among TB patients have been reported

(Kennedy et al. 1996; Mupere et al. 2010). There may also be increased production of pro-inflammatory cytokines resulting in breakdown of body lipids and proteins

(Niyongabo et al. 1999; van Lettow, Fawzi, and Semba 2003) in a patient with already compromised nutrient intake because of reduced appetite. Moreover, wasting itself is a cause of immune-deficiency affecting cellular immunity the key host defense against tuberculosis (Chandra 1991; Fernandes G, Jolly C.A, and Lawrence R.A 2006; Gershwin

M.E, Beach R.S, and Hurley L.S 1985). Thus, body wasting may be a risk factor for tuberculosis and tuberculosis might not only worsen the course of HIV-associated immune-depression but also worsen the HIV-associated wasting because recent reports suggest tuberculosis to be the dominant factor in driving the wasting process in co- infection (Mupere et al. 2010; Paton and Ng 2006). These interrelated effects possibly explain the fact that both tuberculosis and wasting are associated with reduced survival during HIV infection (Nunn et al. 1992; Suttmann et al. 1995).

Early detection and timely management of body wasting is important in prevention of morbidity and mortality in tuberculosis. The assessment of body wasting and management is also important during and after tuberculosis treatment to understand the changes in body composition. However up to date, several prior studies particularly in sub-Saharan Africa that assessed body wasting in tuberculosis used body mass index

(BMI) (Zachariah et al. 2002; Kennedy et al. 1996; Mugusi et al. 2009). Yet BMI is insensitive to body fatness, particularly at low BMI, as well as with above normal muscle development (Kyle, Genton, and Pichard 2002; Kyle, Piccoli, and Pichard 2003). This implies that previous studies (Zachariah et al. 2002; Mugusi et al. 2009) might have

overestimated mortality due to wasting, and might have failed to reveal precise changes in body composition during and after tuberculosis (Kennedy et al. 1996; Ramakrishnan et al. 1961). Moreover, compared to fat mass or weight in itself, fat-free mass body compartment correlates closely with quality of life, physical functioning, and survival

(Wagner, Ferrando, and Rabkin 2000; Mostert et al. 2000). Bioelectrical impedance analysis (BIA) provides a precise and practical method for clinical assessment of fat and fat-free mass (Kyle, Genton, and Pichard 2002; Kyle et al. 2004); however, its application is still limited to research settings that simple and inexpensive methods are needed particularly in sub-Saharan Africa.

Although body wasting and malnutrition play an important role in the clinical course of patients with tuberculosis and HIV and those with dual infection, nutritional status, nutritional intake, and quality of intake are often overlooked in regular clinical practice and in tuberculosis programs. This has led to paucity of information to characterize nutritional intake and quality of intake, and how the intake influence body composition in patients with or without tuberculosis. Characterization of tuberculosis-associated body wasting and nutrient intake not only contributes to the understanding of the pathophysiology but also the management of body wasting.

This dissertation was designed to fill gaps in understanding survival and changes in body composition using precise measures of nutritional status; in understanding nutrient intake

and its association with body wasting and body composition measurements; and how body composition can be assessed using simple existing equations in sub-Saharan.

Goal

The goal of this dissertation was to generate information that could be used to improve survival in tuberculosis patients. This dissertation focused first on understanding simple and inexpensive approaches of assessing body wasting. Second, focused on generating information that could aid in understanding of body wasting and its management, that is whether nutritional factors influence body composition in the face of tuberculosis; and whether body wasting as measured by precise measures of body composition modifies the course of tuberculosis.

Figure 1:1 Working tuberculosis – HIV – Malnutrition Model

Tuberculosis HIV Reduced appetite Immune activation Inadequate dietary intake Immune deficiency Altered metabolism: HIV replication Increased cytokine production Anabolic block

HIV-TB Co-infection Increased energy expenditure Malabsorption Increased inadequate intake

Lipid and protein metabolism Wasting of body composition: Men (Mupere et al. 2010) Women Begin with high fat-free mass Begin with low fat-free mass Begin low with fat mass Begin with high fat mass Marked fat-free mass wasting Preserve fat-free mass wasting Fat mass wasted in proportion with FFM Marked fat mass wasting Low body mass index High body mass index Gender differences in micronutrient deficiency not known?

Wasting modifies: Men Women Survival and effect of HIV Survival and effect of HIV? Changes: Fat-free mass, fat mass, BMI? Changes: Fat-free mass, fat mass, BMI? Differences in dietary Intake? Energy and protein intake in prediction of body composition? Independent effects of HIV and tuberculosis on dietary intake?

BMI = body mass index, FFM = fat-free mass

SPECIFIC AIMS

Chapter 4: Body composition measured with bioelectrical impedance analysis and

anthropometry among HIV positive and HIV negative adults with or without

tuberculosis in urban Kampala, Uganda.

Specific Aim:

To establish whether simple anthropometric measurements can be used to provide

comparable estimates of body composition to those of bioelectrical impedance analysis.

Hypothesis:

Waist circumference, mid-upper-arm circumference, and body mass index provide similar estimates of body composition to those of bioelectrical impedance analysis.

Chapter 5: Indicators of dietary adequacy among HIV positive and HIV negative adults with or without tuberculosis in urban Kampala, Uganda.

Specific Aim:

To evaluate whether simple counting of food items or food groups are indicators of dietary adequacy in a population of HIV positive and HIV negative adults with or without tuberculosis in urban Kampala, Uganda.

Hypothesis:

Simple counting of food items or food groups are predictors of dietary adequacy among

HIV positive and HIV negative adults with or without tuberculosis in urban Kampala,

Uganda.

Chapter 6: Predictors of body composition among HIV positive and HIV negative adults with or without tuberculosis in urban Kampala, Uganda.

Specific Aim:

To establish whether energy and protein intake are predictors of body composition among

HIV positive and HIV negative adults with or without tuberculosis in urban Kampala,

Uganda.

Chapter 7: Body wasting and dietary intake among HIV positive and HIV negative adults with or without tuberculosis in urban Kampala, Uganda.

Specific Aim:

To establish the relationship between body wasting or severity of tuberculosis disease and dietary intake among HIV positive and HIV negative adults with or without tuberculosis in urban Kampala, Uganda.

Hypothesis:

HIV modifies dietary intake among tuberculosis patients.

Dietary intake differs by body wasting.

Chapter 8: Correlates of dietary intake and adequacy among HIV positive and HIV

negative adults with or without tuberculosis in urban Kampala, Uganda.

Specific Aim:

To establish dietary correlates of energy and protein intake, and correlates of nutrient inadequacy among HIV positive and HIV negative adults with or without tuberculosis in urban Kampala, Uganda.

Hypothesis:

Tuberculosis and HIV do not affect energy and protein intake

Chapter 9: Impact of body wasting on survival among adult patients with pulmonary tuberculosis in urban Kampala, Uganda.

Specific Aim:

To establish the effect of body wasting as measured by height-normalized fat mass and lean tissue and body mass indices on survival among adult patients with pulmonary tuberculosis in urban Kampala, Uganda.

Hypothesis:

Body wasting modifies survival among tuberculosis patients in urban Kampala, Uganda.

Chapter 10: Longitudinal changes in body composition during and after tuberculosis treatment in urban Kampala, Uganda.

Specific Aim: To establish the rate of change for fat mass, fat-free mass, and BMI among tuberculosis patients in urban Kampala, Uganda.

Hypothesis:

Body wasting modifies longitudinal changes in body composition during and after tuberculosis treatment.

CHAPTER 2

BACKGROUND

Nutrition and Tuberculosis

The role of nutrition in tuberculosis leads to consideration of three important questions

(Rao K.N and Gopalan C 1966). Does malnutrition predispose to tuberculosis? Does malnutrition influence or modify the course of established infection? Do nutritional factors influence the response to chemotherapy? In the present dissertation, the focus was to understand the influence of body wasting and malnutrition as measured by precise measures of fat and fat-free mass on survival and changes in body composition during and after tuberculosis treatment. Further, to understand the influence of nutritional factors, specifically nutrient intake and its quality on body composition in the presence or absence of tuberculosis.

Effects of nutritional status on tuberculosis

The association between tuberculosis and malnutrition has long been recognized, as malnutrition predisposes people to the development of clinical disease; and tuberculosis often exacerbates malnutrition (Macallan 1999). The host protective immune mechanism of infection with Mycobacterium tuberculosis depends critically on the interaction and cooperation between monocyte-macrophages and T-lymphocytes and their cytokines

(Rook and Hernandez-Pando 1996). Substantial experimental evidence suggests that malnutrition can lead to secondary immunodeficiency that increase the host’s

susceptibility to infection. Severe malnutrition has profound effects on cell-mediated

immunity (Chandra 1991), and both macro- and micronutrient deficiency can influence

susceptibility to tuberculosis in animal models (Bhuyan and Ramalingaswami 1973;

McMurray et al. 1990). Increased risk of tuberculosis can result from alteration in the

individual protective function of, or the interaction between T-lymphocytes and macrophages because of nutritional insult (Chan J et al. 1997). Although a causal relationship has yet to be clearly established in humans, malnutrition may impair host responses in patients with tuberculosis, particularly leading to a vicious cycle of malnutrition and infection (Bhaskaram 1992).

An already malnourished individual with compromised cell mediated immunity is more likely to become infected with tuberculosis, and latent infection is more likely to become active tuberculosis. Among individuals with latent tuberculosis, the occurrences of malnutrition may be an important trigger for active tuberculosis development (Cegielski and McMurray 2004). One longitudinal study of participants in a BCG vaccine trial in the

United States found the incidence of active tuberculosis was 2.2 times higher in children with low subcutaneous fat stores (skin-fold thicknesses between 0 and 4 mm) compared to with those with 10 mm subcutaneous fat (Comstock and Palmer 1966). In a large study in Norway, the incidence of smear-positive and smear-negative tuberculosis declined significantly with increasing BMI in all age groups. New tuberculosis diagnosis was 5 times higher in the lowest BMI group compared with the highest BMI group (Tverdal

1986). Individuals with immuno-suppression have a greater risk of developing clinical

tuberculosis which explains the increased prevalence of tuberculosis in association with

HIV infection.

Effect of tuberculosis on nutritional status

During active tuberculosis, catabolic processes that cause wasting usually begins before the patient is diagnosed; therefore more is known about nutritional status at the time of diagnosis than of the wasting process per se (Macallan 1999). Prior studies have examined the effect of tuberculosis on nutritional state and demonstrated extensive nutritional depletion at the time of diagnosis (Onwubalili 1988; Zacharia R et al. 2002;

Zachariah et al. 2002). In developing countries, the impact of tuberculosis is even more dramatic. For example, one study in Malawi that evaluated 122 patients revealed a reduction in BMI of 20% from 21.7 to 17.3 with 35% reduction in skinfold thickness and

19% reduction in arm muscle circumference (Harries et al. 1988). Similar degrees of wasting have been reported in other populations (Scalcini et al. 1991). Such losses of both fat and fat-free mass (muscle or lean tissue) represent severe wasting and may pose significant impact on morbidity and mortality of their own in addition to that due to tuberculosis disease itself.

Effect of co-infection with HIV and tuberculosis on nutritional status

Although co-infection with HIV and tuberculosis may be thought to introduce an extra dimension to the pathophysiology of wasting, exacerbating the wasting seen in

tuberculosis or HIV infection alone (Macallan 1999; Lucas et al. 1994), recent studies

(Mupere et al. 2010) with a full panel of HIV positive and HIV negative adults with or without tuberculosis suggest gender but HIV as the factor associated with differences in body composition during co-infection. Further, tuberculosis has been suggested as the

dominant factor driving the wasting process at the time of diagnosis (Mupere et al. 2010;

Paton and Ng 2006). Prior studies that reported conflicting results were limited in terms

of sample size, comparison groups, and study populations that were composed of only

men (Paton et al. 1999) (Scalcini et al. 1991).

Possible Mechanism of Wasting or effect of tuberculosis on nutrition

In the 21st century, tuberculosis is still the most frequent underlying cause of wasting

worldwide. Wasting is regarded as the cardinal features of the disease. However, the

pathophysiology of wasting in tuberculosis remains poorly understood (Schwenk A and

Macallan D.C 2000). For any infection, there is a complex interaction between the host

response and the virulence of the organisms, which modulates the overall metabolic

response and the degree and the pattern of tissue loss. In patients with tuberculosis, a

reduction in appetite with eventual reduction in energy intake, nutrient malabsorption,

micronutrient malabsoption, and altered metabolism leads to wasting (Paton et al. 1999;

Macallan et al. 1998). However, factors that make the most dominant contribution in

tuberculosis have been difficult to identify because the wasting phase is rarely observed

and because of the ethical obligation to commence treatment once a diagnosis of

tuberculosis has been made. Observations have been made at the time of diagnosis that

the metabolic rate or resting energy expenditure is increased (Macallan et al. 1998).

However, by analogy with HIV infection where wasting phase has been scrutinized

(Macallan, Noble et al. 1995), increased energy expenditure is unlikely to be the sole driving force behind wasting in the catabolic phase. It is reduced energy intake which may be more likely to be the primary driving force.

In one study in India, patients with pulmonary tuberculosis were compared with malnourished and normally nourished healthy subjects. Whereas protein synthesis

(anabolism) and breakdown (catabolism) in the fasting state were not significantly different between groups, patients with tuberculosis used a larger proportion of proteins

(amino acids) from oral feeding for oxidation and hence for energy production than did either control group. The failure to channel food protein into endogenous protein synthesis has been termed “Anabolic block”. This anabolic block represents one of the mechanism for wasting in tuberculosis and other inflammatory status (Macallan et al.

1998; Paton et al. 2003). Tuberculosis may have a greater catabolic effect than HIV infection where similar studies failed to demonstrate anabolic block at the whole body level (Macallan, McNurlan et al. 1995). The difference in metabolic response may be the consequence of differences in the pattern of cytokine activation between tuberculosis and

HIV disease states.

Anorexia is also a contributing factor for wasting in tuberculosis. In an unselected U.S. cohort of patients diagnosed with tuberculosis, 45% lost weight and 20% had anorexia

(Miller et al. 2000). Increased production of cytokines with lipolytic and proteolytic

activity cause increased energy expenditure in tuberculosis (Verbon et al. 1999). Leptin

may also play an important role in wasting (Sarraf et al. 1997). In a study, malnutrition

has been associated with atypical presentations of tuberculosis (Madebo, Nysaeter, and

Lindtjorn 1997).

Micronutrient malnutrition in tuberculosis

Several micronutrient deficiencies have been described in individuals with tuberculosis

(van Lettow, Fawzi, and Semba 2003) and in those with HIV infection (Semba and Tang

1999; Bogden et al. 1990; Beach et al. 1992; Ullrich et al. 1994; Baum et al. 1991;

Ehrenpreis et al. 1994; Harriman et al. 1989). Several cross-sectional studies suggest that

patients with tuberculosis suffer from deficiencies of vitamin A (Evans and Attock 1971;

Smurova and Prokop'ev 1969), thiamin (Arkhipova O.P 1975), vitamin B6 (Miansikov

V.G 1969),folate (Markkanen et al. 1967; Line et al. 1971), and vitamin E (Panasyuk A.V

et al. 1991). Deficiencies that have been reported to be more prevalent among HIV-

infected adults than in those without HIV infection include vitamin A, vitamin E,

thiamin, riboflavin, vitamin B6, and vitamin C (Beach et al. 1992; Ullrich et al. 1994;

Baum et al. 1991; Ehrenpreis et al. 1994; Harriman et al. 1989). Of these deficiencies,

vitamin A and D have received the most attention in patients with tuberculosis. The

interest in vitamins A and D hinges on the historic use of cod-liver oil as treatment for tuberculosis prior to the era of antibiotics (Whalen C and Semba R.D 2001). Vitamins A,

C, E, B6, and folic acid and minerals zinc, copper, selenium, and iron all have key roles in

metabolic pathways, cellular function, and immune competence (Keusch 1990). The concentration of these may have a role in host defense against tuberculosis (Karyadi et al.

2000). Deficiency of single or multiple nutrients can reduce an individual’s resistance to any infection (Chandra 1991; Keusch 1990; Chandra and Kumari 1994).

Changes in Nutritional Status during Tuberculous Chemotherapy

During drug treatment of active tuberculosis without supplementary nutrition, nutritional status usually improves. This can be attributed to a variety of reasons including improved appetite and food intake, reduced energy/nutrient demands, and improved metabolic efficiency. Most improvements, how, are limited to increase in fat mass (Macallan 1999).

For example, a study by Schwenk et al. (Schwenk et al. 2004) that investigated the changes in fat mass and protein mass (fat-free mass) in 40 patients receiving standard tuberculosis treatment found after six months of treatment, the patients had gained 9.5 ±

8.95% body weight, mainly due to gain in fat mass with no significant change in protein mass suggesting that clinical recovery from tuberculosis does not guarantee protein mass restoration, even though weight gain is significant. This finding may support the idea that protein metabolism continues to be altered even during treatment, and that clinical and functional recovery from tuberculosis lags behind microbial cure. Alternatively, diet during treatment may have been inadequate in relationship to increased requirements during treatment and recovery, thereby limiting development of lean body mass.

Nutritional treatment of tuberculosis

Nutritional supplementation may help to improve outcome in tuberculosis patients. One study in Singapore (Paton et al. 2004), found that nutritional counseling to increase energy intake combined with provision of supplements, when started during the initial phase of tuberculosis treatment, produced a significant increase in body weight, total lean mass, and physical function after six weeks. A large proportion (46%) of the early weight gain comprised lean tissue, confirming the findings that tuberculosis can mount a protein anabolic response on feeding. In the same study, patients in the nutritional supplementation group continued to show a greater increase in body weight than control subjects during later follow-up. However, the pattern changed toward deposition of predominantly fat mass, whereas in the control group, the weight gain comprised fat lean tissue in approximately equal proportions (Paton et al. 2004). However, the changes in lean tissue described above could be an underestimate of the actual improvement in nutritional status, given that feeding initially leads to a loss of extracellular water that accumulates in malnourished individuals, including those with tuberculosis (Paton et al.

1999). Accelerating the recovery of lean tissue might help to restore physical functions more rapidly. Restoration of physical function might help to shorten the convalescent period and facilitate earlier return to productive work (Paton et al. 2004). Early restoration of nutrition could also lead to immunologic changes that could enhance the clearance of mycobacteria and reduce infectiousness of patient.

Vitamins and minerals can play important role in treatment of tuberculosis. In a trial among 110 new cases of active tuberculosis, subjects received tuberculosis chemotherapy alone, or in addition to injectable thiamin, vitamin B6, and vitamin C, or oral multivitamin supplement (Volosevich 1982). All groups receiving any vitamin supplementation had significantly better lymphocyte proliferation responses than the group receiving no supplement. Another trial showed that vitamins C and E were effective in improving immune responses to tuberculosis when given as adjuvant to multidrug tuberculosis therapy (Safarian et al. 1990). The supplementation with vitamin

A and zinc improved the effectiveness of the antituberculosis drugs in the first two months. The improved outcome was indicated by the higher number of patients with sputum negative for bacilli and significantly lower mean lesion area in the lungs (Karyadi et al. 2002).

CHAPTER 3

STUDY METHODS

Study Design

To address the aims of this dissertation, a hybrid of retrospective cohort and cross- sectional designs were employed. The completed five year Household Contact (HHC) study, the completed phase II prednisolone double blind randomized placebo controlled clinical trial (PD), and the ongoing Kawempe Community Health (KCH) study databases were used for the retrospective design. Datasets with relevant key variables from the

HHC, KCH, and PD were created and merged to form one large working database. A total of 753 patients were evaluated for analysis. Of the 753, 314 were enrolled into the

HHC, 344 into the KCH, and 95 into the placebo arm of the prednisolone clinical trial

(Figure 3:1). The datasets for the three studies were first tested for differences in baseline characteristics before combining for analysis. The datasets were different regarding extent of tuberculosis disease on chest x-ray because the phase II prednisolone trial enrolled only HIV-associated tuberculosis patients with CD4 cell count >200 cells/l compared to HHC and KCH studies (Appendix, Table 13:1). One of our interests was to establish the confounding effect of HIV; therefore, we combined all the three datasets for analysis. The BIA data (specifically, fat-free mass and fat mass) were collected during the KCH study only.

The HHC and KCH studies were observational epidemiologic studies; organized and conducted by the Makerere University and Case Western Reserve University TB research collaboration (Uganda-CWRU) that has been ongoing for the last 20 years in Uganda.

The HHC was the initial household contact study from 1995 to 1999 that described the epidemiology of TB in urban Kampala, Uganda (Guwatudde et al. 2003). The KCH is the second phase of the HHC. The KCH phase started in 2002 and is still ongoing (Stein et al. 2005). The KCH was developed specifically to focus on the determinants of host factors associated with primary infection, re-infection, reactivation, and progression of clinical disease and to identify and track individual strains of mycobacterial TB through

Ugandan households and local community. The phase II clinical trial was conducted between 1995 to 2000 to determine whether immunoadjuvant prednisolone therapy in

HIV-infected patients with TB who had CD4(+) T cell counts >/=200 cells/ mu L was safe and effective at increasing CD4(+) T cell counts (Mayanja-Kizza et al. 2005).

Patients were eligible for analysis if they were 18 or more years of age, had baseline measurements, had an HIV test, and were part of one prospective study conducted by the

Uganda-CWRU research collaboration. Adults with a previous history of treated pulmonary TB were excluded in the study.

In a cross-sectional study, 132 participants 18 years or older residing in Kampala district or 20 km from the study site if residence was outside Kampala in Uganda were enrolled.

One participant was excluded from the analysis because of prior TB treatment. The study was conducted at the National TB and Leprosy Program (NTLP) Clinic of the national tertiary teaching hospital complex, Mulago between November 2007 and March 2008. Of the 131 participants who were included in the analysis, 31 were HIV positive with TB and 32 were HIV negative with TB and were recruited at the Mulago NTLP Clinic; 38 were HIV positive without TB and recruited at the Infectious Disease Institute Clinic

(IDI) located 500 meters from the Mulago NTLP Clinic; and 30 HIV negative individuals without TB were enrolled from the community where TB patients resided (Figure 3:2).

The retrospective cohort and cross-sectional designs were chosen in view of their advantages. There was a rich existing database on study participants who fit the needed criteria for a retrospective cohort design. A cross-sectional design could easily be implemented in a short period of time for primary data collection to study associations.

The retrospective cohort and cross-sectional were thus cost effective and took time constraints into consideration.

In a retrospective cohort, the investigator identifies a cohort of individuals based on their characteristics in the past and then reconstructs their past or up to the present time (or occasionally into the future). The ideal concurrent cohort design in which the investigator identifies the cohort on the basis of current characteristics and is then followed forward in time would be expensive and would require a longer time to complete the study. The potential bias of missingness in data anticipated in a retrospective cohort design was assessed for using standard statistical procedures.

In a cross-sectional design, the investigator evaluates for the exposure and outcome of interest at the same one point in time. The disadvantage with a cross-sectional design is that the presence or absence of both exposure and disease/outcome of interest are

determined at the same time in each individual in the study. It is not possible to establish

a temporal relationship between exposure and onset of disease. Therefore, although a

cross-sectional study can be very suggestive of a possible risk factor or factors for a

disease, with limitations in establishing the temporal relationships, the association cannot

reflect a causal relationship.

The institutional review boards at Case Western Reserve University in the United States

and Joint Clinical Research Center in Uganda reviewed the protocol and final approval

was obtained from the Uganda National Council for Science and Technology. All

patients had written informed consent taken to be enrolled in the parent studies. All

participants were given appropriate pre- and post-test HIV counseling and AIDS education.

Measurements

In all the four studies, socio-demographic and clinical information was obtained through standardized interviews and physical examination performed by trained medical officers.

Venous blood was collected for HIV-1 enzyme immunoassay testing and complete blood and differential counts. HIV infection was documented by enzyme-linked immunosorbent assays. None of the HIV positive patients, neither those who were newly identified with

HIV nor those with pre-existing HIV, were on antiretroviral therapy. All participants had posterior-anterior chest X-rays taken at baseline. Expectorated sputum specimens were collected, concentrated, and stained for acid fast bacilli (AFB) with Ziehl-Neelsen stain at

the Wandegeya national reference laboratory in Uganda. AFB smears were reviewed by

trained technicians who graded the smears by the number of acid-fast organisms seen on

the light microscopy according to criteria established by the WHO (International Union

Against Tuberculosis and Lung Disease 1986). Specimens were cultured for

mycobacteria tuberculosis on Lowenstein-Jensen medium slants, incubated at 370C in air and examined weekly until positive or for 8 weeks. Patients with active tuberculosis were treated with standard four-drug chemotherapy for tuberculosis per guidelines of the

Ugandan Ministry of Health.

Nutritional status was assessed using anthropometric measurements such as height and

weight and BIA Detroit, MI, RJL Systems. Body weight was determined to the nearest

0.1 kg using a SECA adult balance, and standing height was determined to the nearest 2 mm. Body-mass index (BMI) was computed using the relationship of weight in kilograms divided by height in meters squared (kg/m2). All BIA measurements were

performed by one trained observer (except measurement for the cross-sectional study that

had two medical officers) using the same equipment and recommended standard

conditions with regard to body position, previous exercise, dietary intake, skin

temperature, and voiding of the bladder were taken into consideration in taking BIA

measurements (Kyle et al. 2004). All BIA measurements during the KCH study were

performed on the day patients were confirmed to have TB disease.

The BIA is a simple, easy, safe, non-invasive technique, that has been recommended for nutritional studies in the clinical setting (Kyle et al. 2004; Kyle, Piccoli, and Pichard

2003) and is a convenient method to determine the lean or fat-free mass and fat body compartments (Kyle, Piccoli, and Pichard 2003; Kyle et al. 2004). Single-frequency BIA was performed at 50 kHz and 800 mA with standard tetrapolar lead placement (Jackson et al. 1988). Before performing measurements on each participant, the BIA instrument was calibrated using the manufacturer’s recalibration device. The resistance and reactance were based on measures of a series circuit (Kotler et al. 1996). BIA measurements were performed in triplicate for each subject. Fat-free mass was calculated from BIA measurements using equations that were previously cross-validated in a sample of patients (white, black and Hispanic) with and without HIV infection (Kotler et al.

1996) and have been applied elsewhere in African studies (Villamor et al. 2006; Shah et al. 2001; Van Lettow et al. 2004). Fat mass was calculated as body weight minus fat-free

mass.

Operational definition

We used BMI and height-normalized indices (adjusted for height2) of body composition

that partition BMI into fat-free mass index (FFMI) and fat mass index (FMI) (Schutz,

Kyle, and Pichard 2002; VanItallie et al. 1990; Kyle, Piccoli, and Pichard 2003) to

establish the body wasting status of participants. The FFMI and FMI have the advantages

of compensating for differences in height and age (Kyle, Genton, and Pichard 2002).

Also, the use of the FFMI and FMI eliminates some of the concerns about differences

between population groups. We defined body wasting as patients having the low fat-free

mass index (FFMI) and the low body fat mass index (FMI) corresponding to WHO BMI

categories for malnutrition as previously reported (Kyle, Piccoli, and Pichard 2003). The

FFMI <16.7 (kg/m2) for men and <14.6 (kg/m2) for women and the FMI <1.8 (kg/m2) for

men and <3.9 (kg/m2) for women corresponds to a BMI of <18.5 kg/m2, the WHO cutoff

for malnutrition (World Health Organ Tech Rep 1995) among adults.

Statistical Analyses

Descriptive statistics

This involved distribution of all variables of interest by computing frequencies and

proportions for categorical variables, means and medians for continuous variables. In

measures analysis, plots of individual participants’ and group mean trajectory response

profiles were performed to assess how BMI, FFMI, and FMI evolved over time and how

measurements made at different time points were related.

Bivariate associations

Chi-square and Fisher’s exact tests were used to assess associations and differences in

proportions between categorical variables. Fisher’s exact test was used where tabular

counts were less than five. Student’s t-test and analysis of variance (ANOVA) were used to compare continuous variables across categories. Variance ratio testing was performed to determine whether to use equal or unequal variances for the Student’s t-test and

Bartlett’s test for equal variances was used to assess equality of variance assumption for

ANOVA. Non-parametric test (Mann-Whitney test) was used where continuous variables were not normally distributed or when sample size distribution was small.

Stratified analysis

Stratified analysis was performed according to key confounding variables such as gender

and HIV to assess for confounding and effect modification.

Univariate and multivariable analyses

Univariate analyses were performed to assess unadjusted associations between the

outcome variables and the main predictor and the potential confounders.

Multivariable analyses were performed using multivariable modeling to estimate adjusted

measures of associations while taking into account multiple confounding variables.

Variables for inclusion in the multivariable models were selected based on biological

plausibility and statistical significance in univariate analysis.

Further analyses were performed according to each specific aim.

Figure 3:1 Retrospective cohort study flow

658 Household Contact 95 Placebo arm Prednisolone

753 Study participants

Exclusion: 6 <18 yrs

Subset with BIA: 311

BMI: wasted No wasting FFMI: Wasted No wasting 310 (42%) 437 (58%) 103 (33%) 208 (67%)

Repeated Measurements: Baseline & Follow-up: BMI at 0, 2, 3, 5, 6, 12, and 24 FFM, FM at 0, 3, 12, and 24

Household contact includes 344 from the initial household contact study and 314 from the Kawempe Community Healthy study. BMI = body mass index, FFM = fat-free mass, FFMI = fat-free mass index, FM = fat mass, and FMI = fat mass index

Figure 3:2 Cross-sectional study flow diagram

137 Screened >18 years Exclusions: 4 Declined 1 Appointment 1 Prior TB Rx 131 included in analysis

32 TB no HIV 30 TB with HIV 38 HIV no TB 31 no HIV no TB

Measurements: 24-Hour Dietary Intake Recall Weight, height, BMI, Fat-free mass, Fat mass Clinical severity of TB disease: TBscore with 13 clinical variables; 0-5 mild and >5 moderate/severe disease

TB = tuberculosis, BMI = body mass index.

CHAPTER 4

BODY COMPOSITION MEASURED WITH BIOELECTRICAL IMPEDANCE

ANALYSIS AND ANTHROPOMETRY AMONG HIV POSITIVE AND HIV

NEGATIVE ADULTS WITH OR WITHOUT TUBERCULOSIS IN KAMPALA,

UGANDA

Abstract

Background Simple and inexpensive method for assessing body composition is lacking in sub-Saharan Africa, a region with high burden of tuberculosis and HIV dual epidemic to ensure early detection and timely management of population groups at risk of fat-free

mass (FFM, lean tissue) wasting associated with poor outcome.

Objective We determined to establish whether existing equations that involve simple

anthropometric measurements of waist circumference (WC), body mass index (BMI), or

mid-upper arm circumference (MUAC) to provide comparable results to those of

bioelectrical impedance analysis (BIA).

Method In a cross-sectional study of 131 participants who were screened for active

tuberculosis and HIV infection, reactance, resistance, height, weight, MUAC, waist and

hip circumferences, and four skinfold thickness (SF): triceps, biceps, subscapular, and

suprailiac were measured. We compared fat mass and FFM as measured by BIA and as

measured by equations that involved waist circumference, BMI, or MUAC using Bland-

Altman plots.

Results On average, the equation that involved WC showed no bias in estimating

comparable FFM to that of BIA among men with tuberculosis regardless of HIV status

while in women, the equation showed no bias among individuals without tuberculosis

regardless of HIV status. The equation showed no bias in estimation of fat mass in

comparison to BIA among men and women without tuberculosis regardless of HIV

status. The equation that involved four SF underestimated FFM and overestimated fat

mass among men and among women regardless of tuberculosis and HIV status. There

was no bias among men and women with tuberculosis regardless of HIV status as regards the equation that involved BMI in estimation of FFM. Concerning fat mass determined by this equation, there was no bias among women without tuberculosis and there was no bias in estimating fat mass among men with tuberculosis regardless of HIV status. There was no bias in fat mass and FFM as determined by the equation that involved MUAC in comparison to BIA among men and women regardless of tuberculosis and HIV status.

Conclusion The existing equations with simple anthropometric measurements that provided comparable results to BIA differed by gender and tuberculosis disease status.

During active tuberculosis, equations that involved MUAC and BMI providing comparable results of body composition to those of BIA among men and among women whereas in the absence of active tuberculosis among men, the equation with MUAC provided comparable results and among women, the equation with WC provided comparable results regardless of HIV status. Further studies are needed to validate the

BIA and the equation with simple anthropometric measurements in assessing body composition in an African population.

Background

Body wasting and malnutrition are associated with tuberculosis, and co-infection with

HIV and tuberculosis may be potentially exacerbate the wasting that occurs in

tuberculosis or HIV infection alone (Macallan 1999; Lucas et al. 1994; Schwenk A and

Macallan D.C 2000; van Lettow, Fawzi, and Semba 2003). In sub-Saharan Africa with

the highest overlapping tuberculosis and HIV infection epidemic, by the time patients

present for registration and treatment, a significant proportion have a profound degree of

body wasting and malnutrition (Niyongabo et al. 1999; Kennedy et al. 1996; Zachariah et

al. 2002). Moreover, progressive loss of body mass in TB and HIV-related wasting are strong risk factors for morbidity, mortality, and impaired physical function (Zachariah et al. 2002; Van Lettow et al. 2004; Tang et al. 2002; Harries et al. 1988). Body weight

compartments of fat and fat-free mass differ in their contribution to body wasting and in

their contribution to clinical benefit. For example, fat-free mass (the lean tissue) is more

closely correlated with morbidity, mortality, quality of life, and physical functioning than

fat mass and body weight itself (Wagner, Ferrando, and Rabkin 2000; Mostert et al.

2000; Heitmann et al. 2000).

Assessment of body composition in tuberculosis and HIV patients in sub-Saharan Africa,

a region with 79% of the overlapping global HIV-tuberculosis disease burden, (Lawn and

Churchyard 2009) is critical to ensure prevention, early detection, and timely management of body wasting. Further, assessment is crucial to understand the

pathophysiology and body compartments involved in body wasting. However, most of

the available reference laboratory-based techniques (Norgan 2005), such as air-

displacement plethysmography (ADP), underwater weighing and dual X-ray absorptiometry (DXA) that provide accurate data require a high level of technical expertise; they are time-consuming, unportable, and expensive for field settings in sub-

Saharan Africa. Bioelectrical impedance analysis (BIA) is an easy, safe, non-invasive,

and convenient method recommended to determine to fat and fat-free mass body

compartments (Kyle, Piccoli, and Pichard 2003; Kyle, Bosaeus, De Lorenzo, Deurenberg,

Elia, Gomez et al. 2004); however, its use in sub-Saharan Africa is still limited to

research settings and validation studies are lacking. Skinfold thickness (SF) measurement

is a simple and inexpensive recommended method (Norgan 2005); however, it requires

multiple measurements and the most widely used Durnin and Womersley equation

(Durnin and Womersley 1974) which was established in a Caucasian sample has been

found to be unsuitable in African populations (Oosthuizen et al. 1997).

To fill the current gap in assessing body composition in routine clinical and

epidemiological settings, and to validate population specific equations for use in sub-

Saharan Africa, we present results showing simple existing equations that involve waist

circumference, body-mass index (BMI), and mid-upper-arm circumference (MUAC) to

provide comparable fat and fat-free mass body composition to that of BIA among HIV

positive and HIV negative adults with/or without tuberculosis in urban Kampala, Uganda.

Methods

Subjects

In a cross-sectional study, 132 participants 18 years or older residing in Kampala district or 20 km from the study site if residence was outside Kampala in Uganda were enrolled.

One participant was excluded from the analysis because of prior tuberculosis treatment.

The study was conducted at the National tuberculosis and Leprosy Program (NTLP)

Clinic of the national tertiary teaching hospital complex, Mulago between November

2007 and March 2008. Of the 131 participants who were included in the analysis, 31

were HIV positive with tuberculosis and 32 were HIV negative with tuberculosis and

were recruited at the Mulago NTLP Clinic; 38 were HIV positive without tuberculosis

and recruited at the Infectious Disease Institute Clinic (IDI) located 500 meters from the

Mulago NTLP Clinic; and 30 HIV negative individuals without tuberculosis were

enrolled from the community where TB patients resided. The institutional review boards

at Case Western Reserve University and Joint Clinical Research Center approved the

study, with final approval by the Uganda National Council for Science and Technology.

All participants provided written informed consent to the study.

All subjects in the study were given appropriate pre- and post-test HIV counseling and

AIDS education. HIV-1 infection was diagnosed on the basis of a positive enzyme-linked immunosorbent assay for HIV-1 antibodies (Recombigen; Cambridge Biotech,

Cambridge, MA). At enrollment, basic demographic information and a medical history

were collected, and a standardized physical examination was conducted by a medical

officer. Active pulmonary tuberculosis was confirmed by sputum smear microscopy and

culture. Patients with active tuberculosis were treated with standard four-drug chemotherapy for tuberculosis per guidelines of the Ugandan Ministry of Health.

Similarly, HIV positive patients who were eligible for antiretroviral therapy were started on treatment and cotrimoxazole prophylaxis at the IDI clinic.

Body composition assessment

From anthropometry

Anthropometric measurements included height, weight, MUAC, waist and hip circumferences, and four SF: triceps, biceps, subscapular, and suprailiac. BMI was computed using the relationship of weight in kilograms divided by height in meters squared (kg/m2). Weight was taken using Hanson digital electronic scales to the nearest

100g. Standing height was measured to the nearest cm using a stadiometer. All

circumferences were measured to the nearest 0.1 cm with a non-elastic metric measuring

tape. The left MUAC was measured over the biceps at the mid-point between acromion

and olecranon with the participant’s arm relaxed. The waist circumference was measured

on a horizontal plane at the narrowest part of the torso, i.e., the smallest horizontal

circumference in the area between the ribs and iliac crest. For measurement of the hip

circumference, the measurer squatted beside the participant to judge the level of

maximum extension of the buttocks. The circumference was measured at this level on a

horizontal plane. The SF measurements were taken on the left side of the body using a

Lange skinfold caliper model 12-1110 to nearest millimeter in each participant. All

reported anthropometric measurement values were the mean of duplicates.

The measurements were made using standard procedures (Lohman T.G, Roche A.F, and

Martorell R 1988; Frisancho A.R 1990) by two trained medical officers who took the

standard history and physical examinations. Technical errors of measurements (TEM)

were computed for each trained officer by measuring height, weight, MUAC, waist and

hip circumferences, and four SF of a set of 6 individuals. The trained officer and the

supervisor made measurements of each individual in succession (reading 1). Serial

second measurements of MUAC were then obtained (reading 2). Intra-subject reliability was tested by performing using the duplicates. Intra-subject differences were calculated in absolute values. The TEM =

(reading 1 reading 2)squared/2 X number of duplicates) and the percent

�reliability∑ (TEM− x 100/overall mean of the measurements) for all measurements of each

trained medical officer were calculated. Based on six repetitions, the TEM and reliability

coefficient were within acceptable limits (Lohman T.G, Roche A.F, and Martorell R

1988; Frisancho A.R 1990) (Appendix, Tables 12:1, 12:2, and 12:3).

The following existing equations validated among Caucasians by Lean et al. (Lean, Han,

and Deurenberg 1996) were used to estimate percentage fat mass. The equations were

developed from different combination of anthropometric measurements including waist

circumference, BMI, and MUAC adjusted for sex and age. We selected these equations

because few simple measurements are needed to estimate fat mass compared to using SF.

Furthermore, they were shown to have good prediction of body density with least bias,

such as the one involving waist circumference in both men and women. Fat-free mass was computed by subtracting fat mass from weight in kilograms.

Percent fat mass = (0.567 x waist circumference) + (0.101 x age) – 31.8 for men

= (0.439 x waist circumference) + (0.221 x age) – 9.4 for women

= (1.33 x BMI) + (0.236 x age) – 20.2 for men

= (1.21 x BMI) + (0.262 x age) – 6.7 for women

= (1.52 x MUAC) + (0.336 x age) – 38.7 for men

= (1.38 x MUAC) + (0.243 x age) – 16.7 for women.

In addition, the widely used Durnin and Womersely equation (Durnin and Womersley

1974) that transforms the log sum of the four SF was used to show its lack of agreement with the BIA body composition.

Body density = 1.1765 – 0744 x log∑(triceps + biceps + subscapular + suprailiac) for men,

= 1.1567 – 0.0717 x log∑(triceps + biceps + subscapular + suprailiac) for women.

Body density is converted into percent body fat using Siri’s equation (Siri W.E 1961).

Percent fat mass = [(4.95/body density) – 4.5] x 100.

From bioelectrical impedance analysis

The single-frequency bioelectrical impedance analyzer (BIA Detroit, MI, RJL Systems) performing at 50 kHz and 800 mA was used for BIA measures with detecting electrodes placed on the wrist and ankle and signal introduction electrodes placed on the first joint of the middle finger and behind the middle toe. Before performing measurements on each subject, the BIA instrument was calibrated using the manufacturer’s recalibration device.

The resistance and reactance were based on measures of a series circuit (Kotler et al.

1996). BIA measurements were performed in duplicate for each subject. The analyzer was calibrated monthly. Fat-free mass was calculated from BIA measurements using equations that were previously cross-validated in a sample of patients (white, black and

Hispanic) with and without HIV infection (Kotler et al. 1996) and have been applied elsewhere in African studies (Shah et al. 2001; Van Lettow et al. 2004; Villamor et al.

2006). Fat mass was calculated as body weight minus fat-free mass.

Analysis

The characteristics of men and women were compared using Wilcoxon-Mann-Whitney test for continuously distributed variables due to lack of normality. Height-normalized indices (adjusted for height2) of fat mass (FMI) and fat-free mass (FFMI) were computed

(Schutz, Kyle, and Pichard 2002; VanItallie et al. 1990; Kyle, Piccoli, and Pichard 2003) and compared between women and men. The FFMI and FMI have the advantages of compensating for differences in height and age (Kyle, Genton, and Pichard 2002). Also,

the use of the FFMI and FMI eliminates some of the differences between population

groups.

Fat mass and fat-free mass measured by BIA were compared with the corresponding

anthropometric measures using Spearman correlation coefficients (r) and the method of

Bland and Altman (Bland and Altman 1986), by plotting the bias (BIA measure –

anthropometric measure) against the average of the measure by BIA and anthropometric.

The average bias was calculated as the mean difference in the BIA and the

anthropometric measures and limits of agreement as mean difference ± 1.96 x S.D. To

evaluate whether bias was constant across all levels of the measure, the least-squares

regression slope of the bias was estimated. A slope difference from zero indicated that the

magnitude of bias depended on the level of the measure. Paired t-tests were used to detect

significant differences in body composition obtained using BIA compared to

anthropometric method. The main hypothesis tested was that there were no differences in

measures of fat mass or fat-free mass by BIA compared to anthropometric method (i.e.,

zero mean bias) using a two-tailed α of 0.05. Stratified analysis was performed according

to gender and HIV status. All analyses were performed using SAS software version 9.2

(SAS Institute, Cary, North Carolina).

Results

Overall men were taller, had more fat-free mass as measured by BIA and equation

involving waist circumference, and more fat mass as estimated using equation that

involve a sum of four SF. However, men had lower BMI, SF, hip circumference, and fat mass as measured by BIA and equation involving waist circumference compared to women regardless of tuberculosis and HIV status (Tables 4:1 and 4:2). Men and women had similar waist circumference and MUAC regardless of tuberculosis and HIV status.

There were strong positive correlation coefficients between the measurements made using BIA and the measurements made using equations that involved waist circumference, BMI, or MUAC as presented in Tables 4:3 to 4:7. The high correlation coefficients indicate strong relationships between the measurements made using BIA and the measurements made using equations that involved waist circumference, BMI, or

MUAC. However, a comparison of differences between the measurements made using

BIA and the measurements made using equations that involve anthropometric measures for individual subjects shows that the high correlation coefficients disguise large inter- method bias and error. For example in table 4:3, whereas BIA and the equation involving sum of four SF correlated well (r = 0.36, p<0.001) for estimation of fat mass, the equation involving sum of four SF overestimated fat mass by a mean of 19.41 with an error of 6.76.

Fat-free mass determined by the equation using waist circumference showed no bias in the total study population (Table 4:3, Appendix Figure 12:1) and no bias among all men

(-0.50 ± 3.53, p=0.27) and among all women (0.31 ± 3.94, p=.52; Table 4:4) compared to

BIA. On average, the equation showed no bias among men with TB regardless of HIV

status while in women, the equation showed no bias among individuals without TB regardless of HIV status. However, fat-free mass was overestimated among HIV negative men with no tuberculosis (-2.42 ± 2.96, p=0.005) and underestimated among HIV positive women with TB (2.45 ± 2.71, p=0.001), but in both the bias did not depend on the level of fat-free mass. The slopes of bias were 0.29 and -0.05, respectively with p- value >0.05 for both (Table 4:4). The equation showed no bias in estimation of fat mass in comparison to BIA among men and women without TB regardless of HIV status although on average there was overestimation of fat mass in the overall population (Table

4:3, Appendix Figure 12:2), in all men and in all women (Table 4:4).

The Durnin and Womersely equation that involved sum of four SF underestimated fat- free mass in the total study population (17.86 ± 7.38, p<0.001) and it overestimates fat mass (-19.41 ± 6.76, p<0.001) compared to BIA (Table 4:3, Appendix Figure 12:3). The bias in fat-free mass and in fat mass measures were depended on the level of fat-free mass and fat mass; the slopes of bias were 0.50 and 0.78, respectively with p<0.001. In general, the Durnin and Womersely equation underestimated fat-free mass and overestimate fat mass among men and among women regardless of TB and HIV status

(Table 4:5, Appendix 12:3 and 12:4).

The fat-free mass determined by the equation using BMI showed no bias in the overall population in comparison to BIA (Table 4:3, Appendix Figure 12:5). Specifically in stratified analysis, there was no bias among men and women with TB regardless of HIV

status (Table 5). Regarding fat mass determined by this equation, there was no bias among women without TB regardless of HIV status (Table 5, Appendix 12:6 fat mass for overall population). The slope (0.05, p>0.05) of the bias for fat mass among HIV positive men with TB was not significant suggesting that the bias in fat mass measure was not dependent on fat mass (Table 5). Thus, the equation had no bias in estimating fat mass in comparison to BIA among men with TB regardless of HIV status.

For the entire study population on average, the equation using MUAC overestimated fat- free mass in comparison to BIA (-1.40 ± 3.39, p<0.001), but the bias in the fat-free measure was not dependent on fat-free mass (Table 2, Appendix Figure 12:7). Similarly in stratified analysis regardless of gender and HIV status, the slopes of the bias for fat- free mass were not significant suggesting lack of dependence for the bias in fat-free mass measure on the fat-free mass (Table 6). One can therefore say that there was no bias in fat-free mass determined by the equation involving MUAC in comparison to BIA among men and women regardless of TB and HIV status. Following the same argument, among men regardless of TB and HIV status, the equation showed no bias in fat mass estimation compared to BIA (Table 6, Appendix Figure 12:8). The equation also showed minimal bias in estimating fat mass among women regardless of TB and HIV status.

Discussion

This is the first cross-sectional study in sub-Saharan Africa to show equations with simple anthropometric measurements that provide comparable results of body

composition among adults to that of BIA. In the present cross-sectional study of 131 participants, we tested existing equations (Lean, Han, and Deurenberg 1996) that use waist circumference, BMI, or MUAC to provide comparable results of body fat and fat- free mass with BIA in a population of HIV positive and HIV negative adults with/or without active TB. The equations that provided comparable results to BIA differed by gender and TB disease status. The equation that involved MUAC or BMI provided comparable results of fat-free mass to that of BIA whereas the equation that involved

MUAC provided comparable results of fat mass among men and among women with TB regardless of HIV status. Among men without TB, the equation that involved MUAC provided comparable results of fat and fat-free mass to that of BIA whereas among women without TB, the equation that involved waist circumference provided comparable results of fat and fat-free mass regardless of HIV status. The equations involved a sum of four SF by Durnin and Womersely (Durnin and Womersley 1974), provide biased estimates of fat and fat-free mass that were not comparable to BIA regardless of gender and TB status.

The findings in this study appear to suggest that active TB, not HIV infection, may be the determining factor to consider in selecting equations that will provide comparable results of fat and fat-free mass to that of BIA. In the face of active TB, equations that involved

MUAC and BMI were the equations that had no or minimal bias in providing comparable results of body composition to those of BIA among men and among women regardless of

HIV status. In the absence of active TB among men, the equation that involved MUAC had no or minimal bias in providing comparable results of body composition to those of

BIA whereas among women, the equation that involved waist circumference had no or minimal bias in providing comparable results regardless of HIV status. To our knowledge, this is the first study in sub-Saharan Africa to provide evidence for selecting appropriate equations that involve simple anthropometric measurements for use in estimating fat and fat-free mass among men and women. Previous studies in Africa

(Oosthuizen et al. 1997; Dioum et al. 2005), have evaluated the Durnin and Womersley

(Durnin and Womersley 1974) equation that involve a sum of four SF with biased results of body composition similar to what has been revealed in the present study. The major strengths of our study stem from the heterogeneity of the study population that involved a full panel of HIV positive and HIV negative adults with/ or without active TB among men and women.

In the present study, equations that use simple anthropometric measurements to provide comparable results of body composition to those of BIA among men and women differed by TB disease status. The equations that involve MUAC or BMI provided comparable results of body composition to those of BIA among men and women with TB whereas the equation with MUAC and the equation with waist circumference provide comparable results among men and among women without TB, respectively regardless of HIV status.

The possible explanation to this gender difference in the face of active TB appears to reflect the effect of TB on the nutritional status of patients with the disease. Further, it reflects the dominant nature of TB disease in inducing the wasting process at the time of diagnosis among patients with HIV associated TB. This finding been has reported previously (Paton and Ng 2006; Mupere et al. 2010), and it is reflected in the results of

the present study. For example in this study, in the presence of active TB both men and women had low comparable BMI and MUAC whereas in the absence of active TB, women had BMI and MUAC of higher magnitude compared to men. Moreover, MUAC has been found to reflect adult nutritional status as defined by BMI (Collins 1996;

Powell-Tuck and Hennessy 2003), and MUAC is effective in the determination of malnutrition among adults in developing countries (Bisai and Bose 2009; James et al.

1994). Further, MUAC is a simpler measure than BMI, requiring a minimum of equipment and in practice has been found to predict morbidity and mortality as accurately as deficits in weights (Briend et al. 1989). The equation with waist circumference providing comparable results among women in the absence of active TB is in line with the known fact that women have a higher content of body fat compared to men (Blaak

2001). Moreover, waist circumference has been reported as sensitive indicator of total and central adiposity fat (Taylor et al. 1998; Flegal et al. 2009).

The interpretation of results in the present study is not without limitations. In this study, we used the BIA method in measurement of body composition, yet it is not of reference standard like the dual-energy x-ray absorptiometry. The BIA prediction method used has not yet been validated in the local population. As a result, findings of body composition may be biased because of variations in hydration across ethnic groups (Kyle, Bosaeus, De

Lorenzo, Deurenberg, Elia, Manuel Gomez et al. 2004). However, the equations that were used in this study were previously cross validated in individuals of different race

(white, black, and Hispanic) among men and women, who were both healthy controls and

HIV-infected patients (Kotler et al. 1996). Moreover, the equations have been used

widely in other studies from Africa with meaningful findings (Shah et al. 2001; Van

Lettow et al. 2004; Villamor et al. 2006; Mupere et al. 2010). Also care was taken in

taking measurements at rest, with proper placement of leads, in participants who had not

exercised or taken alcohol, in participants with voided bladder and ambient temperature.

However, half of the measurements were in patients with underlying illness that may

cause shifts in body water compartments, thereby affecting measurements of fat mass.

Our findings are also limited by the cross-sectional nature of the study that as the body composition change while TB patients receive treatment, selection of appropriate of

equations may change with time. We cannot evaluate such temporal changes in the

present study.

Despite limitations of the present study, findings in this study revealed remarkable gender

in the presence or absence of active TB but not HIV that may influence selection of

appropriate equations with waist circumference, BMI, or MUAC to provide comparable

results of body composition to that of BIA. Further studies are needed to validate the BIA

and the equation with simple anthropometric measurements in assessing body

composition in an African population.

Acknowledgements

We thank all study staff members of the Case Western Reserve University and Makerere

University research collaboration at the Tuberculosis Research Unit in the United States

and in Uganda for their assistance; the faculty of staff at Case Western Reserve

University Department of Epidemiology and Biostatistics for the guidance in analyzing the project; and the Fogarty International Center, for the continued support.

This study was supported in part by the AIDS International Training Research Program,

Fogarty International Center, Grant No. TW000011, based at Case Western Reserve

University, Department of Epidemiology and Biostatistics. This work was part of Ezekiel

Mupere’s PhD thesis at Case Western Reserve University.

Table 4:1 Characteristics of study population with tuberculosis

Characteristics (SD)1 HIV positive with TB HIV negative with TB

Men Men Women Women (n=10) (n=21) (n=18) (n=14) Age, years 30.9 (4.6) 29.2 (5.9) 26.0 (7.3) 26.0 (5.4) Weigh, kg 54.3 (6.1) 46.9 (7.4)b 53.2 (6.5) 49.7 (7.6) Height, cm 171.8 (9.3) 158.7 (6.8)a 171.1 (6.0) 157.7 (12.8)a BMI, kg/m2 18.4 (1.7) 18.6 (3.0) 18.2 (2.0) 20.3 (4.3) MUAC, mm 23.0 (2.2) 23.4 (3.0) 23.7 (1.8) 24.5 (2.5) Triceps ST, mm 6.4 (2.3) 13.2 (6.6)b 6.3 (2.0) 16.6 (5.6)a Biceps ST, mm 3.8 (1.0) 6.6 (3.7) 3.6 (0.9) 6.8 (2.7)a Subscapular ST, mm 8.3 (2.7) 11.7 (4.7)b 8.5 (1.9) 13.2 (5.9)b Sacral ST, mm 5.6 (1.7) 9.2 (5.1) 5.4 (1.6) 9.8 (5.1)a Waist circumference, cm 72.5 (4.7) 70.1 (6.3) 70.0 (3.7) 68.5 (5.2) Hip circumference, cm 84.7 (5.1) 86.5 (7.7) 84.4 (4.7) 90.9 (8.1)b BIA FFM, kg 48.7 (6.0) 36.0 (3.8)a 48.1 (5.5) 37.6 (6.2)a BIA FFMI, kg/m2 16.6 (1.3) 15.4 (0.9)b 16.6 (1.5) 16.1 (0.9) WC FFM, kg 47.5 (4.5) 33.6 (4.0)a 47.6 (5.0) 36.6 (4.7)a WC FFMI, kg/m2 16.1 (1.4) 13.3 (1.6)a 16.2 (1.5) 15.1 (3.2)b BMI FFM, kg 48.0 (4.9) 35.6 (4.1)a 47.7 (4.9) 37.3 (4.8)a BMI FFMI, kg/m2 16.3 (1.2) 14.1 (1.4)a 16.3 (1.3) 15.2 (2.1)b MUAC FFM, kg 50.6 (4.2) 35.9 (3.9)a 49.9 (5.2) 38.0 (4.3)a MUAC FFMI, kg/m2 17.2 (1.5) 14.2 (1.6)a 17.0 (1.4) 15.6 (3.0)b BIA Fat mass, kg 5.2 (1.8) 8.3 (5.2)a 4.8 (2.3) 10.5 (6.6)b BIA FMI, kg/m2 1.8 (0.6) 3.3 (2.1)a 1.6 (0.8) 4.6 (3.6)a WC Fat mass, kg 6.8 (2.1) 13.2 (3.5)a 5.7 (1.8) 13.3 (3.3)a

WC FMI, kg/m2 2.3 (0.6) 5.2 (1.5)a 1.9 (0.6) 5.4 (1.5)a BMI Fat mass, kg 6.3 (1.6) 11.2 (3.7)a 5.5 (2.3) 12.7 (4.1)a BMI FMI, kg/m2 2.2 (0.6) 4.5 (1.6)a 1.9 (0.8) 5.3 (2.4) a MUAC Fat mass, kg 3.7 (2.2) 10.9 (3.8)a 3.4 (2.3) 12.0 (3.7)a MUAC FMI, kg/m2 1.2 (0.7) 4.3 (1.5)a 1.2 (0.8) 4.9 (1.6)a ∑4 SF FFM, kg 23.3 (3.5) 21.2 (4.7)b 22.8 (3.6) 23.3 (4.9) ∑4 SF FFMI, kg/m2 7.9 (1.0) 8.4 (1.9) 7.8 (1.1) 9.6 (2.5)b ∑4 SF Fat mass, kg 31.0 (2.7) 25.7 (2.7) 30.4 (3.0) 26.7 (3.0)b ∑4 SF FMI, kg/m2 10.5 (0.8) 10.2 (1.1) 10.4 (1.0) 10.9 (2.0)

ªp-value < 0.001, bp-value < 0.05; p-values were obtained with Mann-Whitney test and were comparisons between men and women. 1Values are means ± standard deviation (SD). TB = Tuberculosis disease, WC = waist circumference, MUAC = mid-upper arm circumference, SF = Skinfold thickness, BMI, body mass index, FFM = Fat-free mass, FFMI = Fat-free mass index, and FMI = Fat mass index.

Table 4:2 Characteristics of study population without tuberculosis

HIV positive no TB HIV negative no TB 1 Characteristics (SD) Men Women Women (n=17) (n=21) Men (n=16) (n=14) Age, years 34.7 (6.5) 29.7 (8.4)b 22.4 (3.2) 24.3 (5.4) Weigh, kg 61.5 (6.3) 57.7 (9.4) 60.5 (6.2) 60.7 (8.7) Height, cm 170.5 (8.2) 155.0 (5.6)a 167.4 (6.6) 159.7 (6.0)b BMI, kg/m2 21.2 (2.2) 24.2 (4.6)b 21.6 (2.3) 23.7 (2.8)b MUAC, mm 27.7 (2.0) 28.5 (3.9) 27.9 (2.1) 29.0 (3.0) Triceps ST, mm 9.8 (7.2) 20.1 (7.2)a 10.5 (4.4) 23.2 (5.7)a Biceps ST, mm 4.5 (2.2) 9.9 (4.4)a 5.2 (2.1) 12.1 (3.9)a Subscapular ST, mm 11.0 (3.8) 19.2 (7.1)a 12.9 (4.2) 17.8 (4.4)a Sacral ST, mm 8.9 (4.2) 15.1 (6.9)b 9.9 (3.9) 16.0 (4.0)a Waist circumference, cm 77.6 (3.5) 77.6 (8.8) 74.6 (5.4) 76.0 (6.1) Hip circumference, cm 91.1 (5.6) 97.0 (9.9)b 90.9 (5.0) 99.3 (6.7)b BIA FFM, kg 51.2 (6.8) 38.2 (3.8)a 49.9 (5.2) 41.3 (6.0)a BIA FFMI, kg/m2 18.3 (1.2) 16.6 (1.2)a 18.6 (1.4) 16.6 (1.1)a WC FFM, kg 52.1 (4.6) 39.4 (5.2)a 52.3 (3.9) 42.7 (5.0)a WC FFMI, kg/m2 18.0 (1.6) 16.5 (2.5) 18.7 (1.5) 16.7 (1.6)b BMI FFM, kg 51.8 (4.6) 39.7 (4.0)a 51.5 (3.8) 43.2 (4.7)a BMI FFMI, kg/m2 17.8 (1.4) 16.6 (1.9)b 18.5 (1.2) 16.9 (1.2)b MUAC FFM, kg 52.5 (5.3) 40.1 (4.4)a 53.1 (3.7) 42.7 (4.6)a MUAC FFMI, kg/m2 18.1 (1.6) 16.7 (2.1) 19.0 (1.6) 16.7 (1.3)a BIA Fat mass, kg 9.1 (3.2) 17.8 (8.2)b 7.9 (2.5) 18.2 (5.4)a BIA FMI, kg/m2 3.2 (1.2) 7.6 (3.7)a 2.9 (1.0) 7.1 (2.1)a WC Fat mass, kg 10.0 (2.4) 18.3 (5.2)a 7.5 (2.2) 18.0 (4.1)a

WC FMI, kg/m2 3.5 (0.8) 7.7 (2.4)a 2.7 (0.8) 7.0 (1.4)a

BMI Fat mass, kg 10.3 (2.7) 18.0 (6.3)a 8.2 (2.5) 17.5 (4.3)a BMI FMI, kg/m2 3.6 (1.0) 7.6 (2.9)a 3.0 (1.1) 6.8 (1.6) a MUAC Fat mass, kg 9.6 (2.9) 17.6 (5.7)a 6.6 (2.6) 18.0 (4.6)a MUAC FMI, kg/m2 3.3 (1.1) 7.4 (2.7)a 2.4 (0.9) 7.0 (1.7)a ∑4 SF FFM, kg 28.2 (3.5) 28.3 (6.1) 27.0 (3.1) 30.4 (5.4)b ∑4 SF FFMI, kg/m2 9.7 (1.3) 11.9 (2.9)b 9.7 (1.3) 11.9 (1.8)a ∑4 SF Fat mass, kg 33.9 (3.4) 29.4 (3.3)a 32.7 (2.6) 30.3 (3.4) ∑4 SF FMI, kg/m2 11.7 (1.1) 12.3 (1.8) 11.7 (1.0) 11.9 (1.0)

Table 4:3 Comparison fat-free mass and fat mass as measured by BIA and by

equations that involved anthropometry measurements for all participants (n=131)

Slope, *p-value 95% limits of Characteristic Correlation bias (se) dBias (SD) agreement

Waist FFm, kg 0.90a -0.04 (0.04) -0.07 (3.77) -7.52, 7.38 0.270

Fat mass, kg 0.83a 0.21 (0.04)a -1.48 (3.31) -7.97, 5.01 0.002 4 Skinfolds FFM, kg 0.43a 0.50 (0.10)a 17.86 (7.38) 3.40, 32.32 <0.001 Fat mass, kg 0.36a 0.78 (0.11)a -19.41 (6.76) -32.66, -6.16 <0.001 BMI FFM, kg 0.93a 0.06 (0.03) -0.50 (2.87) -6.13, 5.13 0.356

Fat mass, kg 0.91a 0.18 (0.03)a -1.05 (2.32) -5.60, 3.50 <0.001 MUAC FFM, kg 0.91a -0.01 (0.04) -1.40 (3.39) -8.04, 5.24 <0.001 Fat mass, kg 0.87a 0.09 (0.04)b -0.15 (2.98) -5.99, 5.69 0.002 ªp-value < 0.001, bp-value < 0.05. Correlations are spearman correlation coefficients. dBias calculated as the mean difference in the BIA and anthropometric measures. 95% limits of agreement calculated as mean difference ± 1.96 x S.D. *P-value for difference obtained by paired t-test. SD = standard deviation, MUAC = mid-upper arm circumference, BMI = body mass index, and FFM = Fat-free mass.

Table 4:4 Comparison fat-free mass and fat mass measured by BIA and by equation with waist circumference

Slope, 95% limits *p- of value Characteristic Correlation bias (se) dBias (SD) agreement

All men FFM, kg 0.81a 0.17 (0.09) -0.50 (3.53) -7.42, 6.42 0.270 Fat mass, kg 0.82a 0.20 (0.06)a -0.69 (1.66) -3.94, 2.56 0.002 HIV+TB+ men FFM, kg 0.81a 0.31 (0.14) 1.26 (2.57) -3.78, 6.30 0.156 Fat mass, kg 0.68b -0.05 (0.22) -1.60 (1.15) -3.85, 0.65 0.002 HIV-TB+ men FFM, kg 0.88a 0.08 (0.11) 0.58 (2.35) -4.03, 5.19 0.307 Fat mass, kg 0.81a 0.28 (0.16) -0.93 (1.50) -3.87, 2.01 0.017 HIV+TB- men FFM, kg 0.78a 0.43 (0.20)b -0.88 (4.68) -10.05, 8.29 0.448 Fat mass, kg 0.77a 0.34 (0.11)a -0.97 (1.97) -4.83, 2.89 0.059 HIV-TB- men FFM, kg 0.87a 0.29 (0.16) -2.42 (2.96) -8.22, 3.38 0.005 Fat mass, kg 0.87a 0.18 (0.10) 0.44 (1.23) -1.97, 2.85 0.173 All women FFM, kg 0.75a -0.13 (0.09) 0.31 (3.94) -7.41, 8.03 0.518 Fat mass, kg 0.89a 0.54 (0.06)a -2.17 (4.14) -10.28, 5.94 <0.001 HIV+TB+ women FFM, kg 0.79a -0.05 (0.17) 2.45 (2.71) -2.86, 7.76 0.001 Fat mass, kg 0.81a 0.43 (0.11)a -4.97 (2.62) -10.11, 0.17 <0.001

HIV- TB+ women

FFM, kg 0.54b 0.35 (0.31) 0.98 (5.41) -9.61, 11.59 0.508 Fat mass, kg 0.66b 0.78 (0.20)a -2.85 (4.91) -12.47, 6.77 0.049

HIV+TB- Women FFM, kg 0.73a -0.35 (0.15)b -1.18 (3.20) -7.45, 5.09 0.108 Fat mass, kg 0.95a 0.49 (0.09)a -0.47 (3.84) -8.00, 7.06 0.579

HIV- TB- women FFM, kg 0.71b 0.19 (0.18) -1.36 (3.41) -8.04, 5.32 0.159 Fat mass, kg 0.72b 0.37 (0.20) 0.19 (3.17) -6.00, 6.38 0.826

ªp-value < 0.001, bp-value < 0.05. Correlations are spearman correlation coefficients. dBias calculated as the mean difference in the BIA and anthropometric measures. 95% limits of agreement calculated as mean difference ± 1.96 x S.D. *P-value for difference obtained by paired t-test. SD = standard deviation, TB- = no TB disease, TB = Tuberculosis disease, HIV- = HIV negative, HIV+ = HIV positive, FFM = Fat-free mass, and FM = Fat mass. .

Table 4:5 Comparison fat-free mass and fat mass measured by BIA and by equation with sum of 4 skinfolds

Slope, 95% limits *p- of value Characteristic Correlation bias (se) dBias (SD) agreement

All men FFM, kg 0.69a 0.42 (0.11)a 24.03 (4.4) 15.5, 32.6 <0.001 Fat mass, kg 0.89a -0.03 (0.10) -25.23 (2.23) -29.6, -20.9 <0.001 HIV+TB+ men FFM, kg 0.67b 0.57 (0.17)a 25.45 (3.39) 18.8, 32.1 <0.001 Fat mass, kg 0.70b -0.42 (0.19) -25.79 (1.48) -28.7, -22.9 <0.001 HIV-TB+ men FFM, kg 0.85a 0.44 (0.12)a 25.32 (2.86) 19.7, 30.9 <0.001 Fat mass, kg 0.81a -0.28 (0.15) -25.67 (1.65) -28.9, -22.4 <0.001 HIV+TB- men FFM, kg 0.52b 0.83 (0.27)a 22.97 (5.94) 11.3, 34.6 <0.001 Fat mass, kg 0.50b -0.07 (0.28) -24.83 (3.14) -31.0, -18.7 <0.001 HIV-TB- men FFM, kg 0.64b 0.59 (0.24)b 22.82 (4.05) 14.9, 30.8 <0.001 Fat mass, kg 0.71b -0.02 (0.23) -24.81 (2.02) -28.8, -20.9 <0.001 All women FFM, kg 0.68a -0.27 (0.10)a 12.48 (4.79) 3.1, 21.9 <0.001 Fat mass, kg 0.86a 0.79 (0.06)a -14.34 (5.08) -24.3, -4.4 <0.001 HIV+TB+ women FFM, kg 0.76a -0.25 (0.18) 14.88 (3.26) 8.5, 21.3 <0.001 Fat mass, kg 0.71a 0.70 (0.12)a -17.40 (3.33) -23.9, -10.9 <0.001

HIV-TB+ women

FFM, kg 0.44 0.33 (0.36) 14.37 (6.08) 2.5, 26.3 <0.001 Fat mass, kg 0.59b 0.93 (0.26)a -16.23 (5.61) -27.2, -5.2 <0.001 HIV+TB- Women FFM, kg 0.52b -0.56 (0.19)a 9.88 (4.53) 1.0, 18.8 <0.001 Fat mass, kg 0.92a 0.86 (0.09)a -11.53 (5.20) -21.7, -1.3 <0.001 HIV-TB- women FFM, kg 0.77b 0.12 (0.16) 10.91 (3.6) 4.7, 17.1 <0.001 Fat mass, kg 0.77b 0.48 (0.16)a -12.08 (3.08) -18.1, -6.0 <0.001 ªp-value < 0.001, bp-value < 0.05. Correlations are spearman correlation coefficients. dBias calculated as the mean difference in the BIA and anthropometric measures. 95% limits of agreement calculated as mean difference ± 1.96 x S.D. *P-value for difference obtained by paired t-test. SD = standard deviation, TB- = no TB disease, TB = Tuberculosis disease, HIV- = HIV negative, HIV+ = HIV positive, FFM = Fat-free mass, and FM = Fat mass.

Table 4:6 Comparison fat-free mass and fat mass measured by BIA and by equation with BMI

Slope, 95% limits *p- of value Characteristic Correlation bias (se) dBias (SD) agreement

All men FFM, kg 0.87a 0.20 (0.07)b -0.35 (2.90) -6.03, 5.33 0.356 Fat mass, kg 0.86a 0.04 (0.05) -0.85 (1.34) -3.48, 1.78 <0.001 HIV+TB+ men FFM, kg 0.88a 0.20 (0.10) 0.77 (1.84) -2.84, 4.38 0.218 Fat mass, kg 0.85b 0.05 (0.16) -1.11 (0.08) -2.67, 0.46 0.002 HIV-TB+ men FFM, kg 0.89a 0.11 (0.11) 0.42 (2.22) -3.93, 4.77 0.429 Fat mass, kg 0.75a 0.00 (0.18) -0.77 (1.68) -4.06, 2.52 0.068

HIV+TB- men FFM, kg 0.85a 0.41 (0.15)b -0.58 (3.89) -8.20, 7.04 0.547 Fat mass, kg 0.92a 0.17 (0.07)b -1.28 (1.27) -3.77, 1.21 0.001 HIV-TB- men FFM, kg 0.82a 0.31 (0.13)b -1.66 (2.53) -6.62, 3.30 0.019 Fat mass, kg 0.81a 0.01 (0.10) -0.33 (1.16) -2.60, 1.94 0.279 All women FFM, kg 0.83a 0.01 (0.07) -0.63 (2.85) -6.22, 4.96 0.067 Fat mass, kg 0.96a 0.37 (0.03)a -1.22 (2.92) -6.94, 4.50 0.001 HIV+TB+ women FFM, kg 0.76a -0.08 (0.15) 0.44 (2.49) -4.44, 5.32 0.432 Fat mass, kg 0.90a 0.38 (0.09)a -2.96 (2.37) -7.61, 1.69 <0.001

HIV-TB+

women FFM, kg 0.68b 0.30 (0.19) 0.30 (3.73) -7.01, 7.61 0.766

Fat mass, kg 0.96b 0.50 (0.09)a -2.16 (3.15) -8.33, 4.01 0.023 HIV+TB-

Women FFM, kg 0.87a -0.06 (0.11) -1.51 (1.87) -5.18, 2.16 0.001 Fat mass, kg 0.98a 0.28 (0.05)a -0.14 (2.36) -4.77, 4.49 0.789 HIV-TB- women FFM, kg 0.81a 0.26 (0.15) -1.87 (2.96) -7.67, 3.93 0.035 Fat mass, kg 0.78b 0.30 (0.15) 0.69 (2.55) -4.31, 5.69 0.329

Table 4:7 Comparison fat-free mass and fat mass measured by BIA and by equation with MUAC

Slope, *p- 95% limits of value Characteristic Correlation bias (se) dBias (SD) agreement

All men FFM, kg 0.80a 0.21 (0.08)b -2.03 (3.32) -8.54, 4.48 <0.001 Fat mass, kg 0.82a -0.08 (0.07) 0.83 (2.01) -3.11, 4.77 0.002 HIV+TB+ men FFM, kg 0.70b 0.36 (0.16) -1.83 (2.86) -7.44, 3.78 0.073

Fat mass, kg 0.67b -0.07 (0.27) 1.49 (1.37) -1.20, 4.18 0.007 HIV-TB+ men FFM, kg 0.82a 0.04 (0.12) -1.74 (2.59) -6.82, 3.34 0.011 Fat mass, kg 0.65b 0.01 (0.22) 1.39 (2.06) -2.65, 5.43 0.011 HIV+TB- men FFM, kg 0.79a 0.26 (0.16) -1.30 (3.97) -9.08, 6.48 0.195 Fat mass, kg 0.83a 0.12 (0.12) -0.55 (1.85) -4.18, 3.08 0.236 HIV-TB- men FFM, kg 0.64b 0.39 (0.20) -3.25 (3.53) -10.17, 3.67 0.002 Fat mass, kg 0.70b 0.11 (0.17) 1.26 (1.89) -2.44, 4.96 0.018 All women FFM, kg 0.77a 0.05 (0.09) -0.85 (3.37) -7.46, 5.76 0.038 Fat mass, kg 0.91a 0.38 (0.05)a -1.01 (3.41) -7.69, 5.67 0.016 HIV+TB+ women FFM, kg 0.70a -0.02 (0.18) 0.17 (2.74) -5.20, 5.54 0.781 Fat mass, kg 0.72a 0.36 (0.12)b -2.69 (2.69) -7.96, 2.58 <0.001

HIV-TB+ women

FFM, kg 0.52 0.45 (0.28) -0.35 (5.04) -10.23, 9.53 0.797 Fat mass, kg 0.82b 0.67 (0.18)b -1.51 (4.50) -10.33, 7.31 0.232

HIV+TB- Women FFM, kg 0.80a -0.16 (0.13) -1.87 (2.28) -6.34, 2.60 0.001 Fat mass, kg 0.99a 0.38 (0.05)a 0.22 (2.89) -5.44, 5.88 0.730

HIV-TB- women FFM, kg 0.77b 0.29 (0.17) -1.35 (3.38) -7.97, 5.27 0.159 Fat mass, kg 0.79a 0.22 (0.18) 0.18 (2.95) -5.60, 5.96 0.824

ªp-value < 0.001, bp-value < 0.05. Correlations are spearman correlation coefficients. dBias calculated as the mean difference in the BIA and anthropometric measures. 95% limits of agreement calculated as mean difference ± 1.96 x S.D. *P-value for difference obtained by paired t-test. SD = standard deviation, TB- = no TB disease, TB+ = Tuberculosis disease, HIV- = HIV negative, HIV+ = HIV positive, FFM = Fat-free mass, and FM = Fat mass.

CHAPTER 5

INDICATORS OF DIETARY ADEQUACY AMONG HIV POSITIVE AND HIV

NEGATIVE ADULTS WITH OR WITHOUT TUBERCULOSIS IN URBAN

KAMPALA, UGANDA

Abstract

Objective To assess nutritional adequacy of dietary intake and validity of simple low-

cost methods to evaluate nutritional adequacy of diets consumed among HIV positive and

HIV negative adults with or without tuberculosis in urban Uganda

Methods In a cross-sectional study of 131 HIV positive and HIV negative adults with or

without tuberculosis, 24-hour dietary intake recall was assessed. Two different dietary

diversity indices were created: food variety score (FVS), a simple count of items, and diet

diversity score (DDS), a count of food groups. Mean adequacy ratio (MAR) of intake to

recommended intake (each truncated at one) of energy and ten nutrients, was calculated

as an indicator of nutrient adequacy.

Results All participants (100%) consumed at least cereals, roots, and tubers, and 90% consumed vegetables not rich in vitamin A such tomatoes and onions while only 45% consumed vitamin-A-rich fruits and vegetables, and only 15% consumed eggs. The mean

FVS and DDS for the study population were low 8.1 ± 2.8 and 4.7 ± 1.4, respectively.

Both men and women regardless of tuberculosis and HIV status, had carbohydrate and ascorbic acid deficiency in the range of 0 to 30% whereas other nutrient intakes including energy, protein, dietary fiber, calcium, magnesium, zinc, iron, vitamin A, vitamin D, and folate had deficiencies ranging 25% to 100%. When a MAR of 65% was used as a cut-off

point for nutrient adequacy, it was found that FVS must be 9 or more and DDS must at

least 5. Among women, both FVS and DDS had a high ability to identify participants

with an inadequate or adequate diet while among men FVS had a high ability to identify

individuals with inadequate diet but low ability to identify those with adequate diet, DDS

had low ability to identify individuals with inadequate diet but had a high ability to identify those with adequate diet.

Conclusion The dietary consumption in this study population was monotonous, rich in carbohydrates and deficient in nutrients regardless of gender, tuberculosis, and HIV status. The ability of FVS and DDS indices to identify individuals with inadequate or adequate diet consumption differed by gender. The FVS was a better predictor of nutritional adequacy among women whereas DDS was among men. Validation in follow- up studies and in different regions is needed.

Background

Micronutrient malnutrition remains one of the largest nutritional problems worldwide,

affecting people in both developed and developing countries (WHO 2002). Among

individuals with tuberculosis and among individuals with HIV infection, a number of

micronutrients deficiencies have been described (van Lettow, Fawzi, and Semba 2003).

Moreover, micronutrient deficiencies not only impair host immune functions, but may

also affect efficacy of tuberculosis drugs (Thurnham 2004), and impair weight gain

which may be important to survival (Paton et al. 2004). The deficiencies may be largely

due to a habitually low dietary consumption of micronutrients in relation to requirements

or due to anorexia (Hopewell P.C. 1994), impaired absorption of nutrients or increased

catabolism.

Dietary diversification is central and sustainable to food based intervention strategies for

combating multiple micronutrient deficiencies and total energy intake in developing

countries (Tontisirin, Nantel, and Bhattacharjee 2002). Among the foremost steps and

important element in planning dietary diversification strategies is dietary assessment.

However, validity studies for simple low-cost assessment methods to evaluate the

nutritional adequacy of diets across different cultural settings, disease conditions, and age

groups are still lacking in sub-Saharan Africa where there is a high burden of multiple

micronutrient deficiencies and tuberculosis with HIV co-infection. The present cross- sectional study was conducted to fill this gap by assessing adequacy of dietary intake and

the relationship between food variety or dietary diversity scores and the nutritional adequacy of the diet among HIV positive and HIV negative adults with or without tuberculosis in urban Kampala, Uganda.

Methods

In a cross-sectional study, 132 participants with age 18 or more years residing in

Kampala district or 20 km from the study site if residence was outside Kampala in

Uganda were enrolled. Data collection was conducted between November 2007 and

March 2008, a period that coincides with harvesting and light rains in November and

December and dry season in January and February. One participant was excluded from the analysis because of prior tuberculosis treatment. The study was conducted at the

National Tuberculosis and Leprosy Program (NTLP) Clinic of the national tertiary teaching hospital, Mulago complex. Of the 131 participants who were included in the analysis, 63 were tuberculosis patients who were recruited at the Mulago NTLP Clinic;

38 were HIV positive patients without TB and recruited at the Infectious Disease Institute

Clinic (IDI) located 500 meters from the Mulago NTLP Clinic; and 30 were HIV negative individuals without tuberculosis from the community where enrolled tuberculosis patients resided. The institutional review boards at Case Western Reserve

University and Joint Clinical Research Center approved the study, with final approval by the Uganda National Council for Science and Technology. All participants provided written informed consent to the study.

All subjects in the study were given appropriate pre- and post-test HIV counseling and

AIDS education. HIV-1 infection was diagnosed on the basis of a positive enzyme-linked immunosorbent assay for HIV-1 antibodies (Recombigen; Cambridge Biotech,

Patients with active TB were treated with standard four-drug chemotherapy for tuberculosis per guidelines of the Ugandan Ministry of Health. Similarly, HIV positive patients eligible for antiretroviral therapy were started on treatment and cotrimoxazole prophylaxis at the IDI clinic. Anthropometric measurements included height and weight.

Body-mass index (BMI) was computed using the relationship of weight in kilograms

divided by height in meters squared (kg/m2). Weight was taken using Hanson digital

electronic bathroom weighing scales to the nearest 100g. Height was measured to the

nearest mm by calibrated standing height using a stadiometer. All anthropometric

measurement values were the mean of duplicates.

The dietary intake assessment was made using a single 24-hour dietary recall with open

ended questions. The reference period for the 24-hour recall was the day prior to the day of the interview. In all instances, the interview was held only with the interviewee, no one

else was present except for children. The questionnaire was pre-tested by administering it

to 8 individuals selected randomly from the neighboring community to the study site. The

assessment was conducted by four trained study nursing staffs and continuously

supervised by a nutritionist using local food photographs, portion-size images, and

volumetric vessels to increase the accuracy of the recall. The nutritive value of raw ingredients was computed using the East African food composition table database whose database was imported into the NutriSurvey software (http:www.nutrisurvey.de) to ease the computations. The database used was predominantly for local Ugandan diet. When the East African food composition table was found deficient in certain food items, the

United States Department of Agriculture database and the African composition table were used.

We evaluated the dietary adequacy in this study population basing methods that have been described previously (Hatloy, Torheim, and Oshaug 1998). The dietary diversity score (DDS) was defined as the number of food groups consumed over a period of 24 hours. The diet was classified according to nine food groups as recommended by Food and Agricultural Organization (FAO) (Table 5:3). Other remaining food items such as tea, sugar, salt, sweets, spices, commercial energy drinks, and alcohol were not used in the DDS and food variety score (FVS) calculations. The FVS was defined as the number of food items consumed over a 24-hour period, from a possible total of 127 Items. The possible total (n=127) reflects all the difference types of food items eaten by this sample population in a 24-hour period.

To estimate the nutrient adequacy of the diet, we calculated the nutrient adequacy ratio

(NAR %) for 10 micronutrients, energy, protein, fat, carbohydrate, and dietary fiber. The

NAR for a given nutrient is the ratio of a participant’s intake to the daily recommended

allowance for the participant’s sex. The Food and Agriculture Organization/World Health

Organization 2002 Human Vitamin and Mineral requirements (FAO/WHO 2002) were

used for vitamin A, vitamin B6, vitamin C, thiamin, riboflavin, folate, magnesium,

calcium, iron, and zinc whereas energy, fat, protein, carbohydrate, and dietary fiber; the

Panel on Macronutrients, Panel on the Definition of Dietary Fiber, Subcommittee on

Upper Reference Levels of Nutrients, Subcommittee on Interpretation and Uses of

Dietary Reference Intakes Food and Nutrition Board (Dietary Reference Intake for

Energy, Carbohydrate, Fiber, Fat, Fatty Acids, Cholesterol, Protein, and Amino Acids

(Macronutrients). A report of the Panel on Macronutrients, Subcommittees on Upper

Reference Levels of Nutrients and Interpretation and Uses of Dietary Reference Intakes, and the Standing Committee on the Scientific Evaluation of Dietary Reference Intakes

2005) was used (Table 5:4). The recommendation for adequate intake was used for vitamin D (Atkinson and Ward 2001). In the case of iron and zinc, the category for

moderate bioavailability was used. The mean adequacy ratio (MAR %) was calculated as

a measure of the adequacy of the overall diet, where MAR is the sum of each NAR

(truncated at 100%) divided by the number of nutrients (excluding energy, fat, protein,

carbohydrates, and dietary fiber) (Madden, Goodman, and Guthrie 1976; Hatloy,

Torheim, and Oshaug 1998).

MAR (Mean Adequacy Ratio) = ∑NAR (each truncated 100%)

Number of nutrients

NAR was truncated at 100% so that a nutrient with a high NAR could not compensate for

a nutrient with a low NAR. For both NAR and MAR a value of 100% is the ideal since it

means that the intake is the same as the requirement.

Analytic strategy

All study participants in the analysis were categorized into 4 mutually independent

groups: HIV positive patients with and without tuberculosis disease, HIV negative

patients with and without tuberculosis disease. Spearman’s rank correlation tests were

performed between NARs or MAR and FVS or DDS. Measures of central tendency and

variability were compared between men and women across 4 mutually exclusive groups

using using Wilcoxon-Mann-Whitney test for average age, weight, height, BMI, FVS,

DDS, and MAR. A non-parametric test was used because of small size issues in

subgroups and nearly all parameters were not normally distributed even after log

transformations. We evaluated FVS and DDS for sensitivity and specificity with MAR as

the ‘gold standard’ of nutritionally adequate intake. Sensitivity is the proportion of

positives that are correctly identified by the test, while specificity is the proportion of

negatives that are correctly identified by the test (Altman D.G 1997). Positive is defined

as FVS or DDS below a given cut-off point. Different cut-off points were tested to find the levels of FVS and DDS that would give high sensitivity without losing too much specificity. Those with a nutritionally inadequate diet, defined as MAR below a certain cut-off point, and FVS or DDS below the cut-off point were defined as true positives.

Those with a nutritionally adequate diet, or MAR greater than the cut-off point, and FVS

or DDS above the cut-off were defined as true negatives. We used linear regression to estimate MAR scores for different levels of FVS and DDS. All analyses were performed using SAS version 9.2 Cary software, (North Carolina SAS Institute Inc 2004) whereas the receiver operator curves plotted in Microsoft Excel 2007.

Results

Of the 131 participants analyzed, 31 were HIV positive with tuberculosis, 32 were HIV negative with tuberculosis, 38 were HIV positive without tuberculosis, and 30 were HIV negative without tuberculosis. Overall men and women in the study population had similar age except among HIV positive individuals without tuberculosis where men were significantly older than women. The average height was significantly higher among men compared to women regardless of tuberculosis and HIV status (Table 5:1 and 5:2).

A total of 127 different food items were eaten by all adults (n=131) in the study population, corresponding to a theoretical maximum of 127 FVS. The mean FVS was 8.1

± 2.8 with a minimum score of 1.0 and maximum of 15.0. The mean DDS value for the total sample was 4.7 ± 1.4 with a minimum of 1.0 and maximum of 8.0. There were no significant differences in average FVS, DDS, and MAR between men and women regardless of tuberculosis and HIV status except for FVS among HIV negative individuals without tuberculosis; women had significantly lower average FVS compared to men (Table 5:1 and 5:2). Of note, women had lower magnitudes of FVS, DDS, and

MAR compared to men regardless of tuberculosis and HIV status.

All participants in the study sample had consumed some kind of cereal, roots, or tubers,

particularly local green plantain, maize meal, sweet potatoes, or rice (Table 5:3). Food

items that were consumed by more than 80% of the sample were other vegetables not rich

in vitamin A including tomatoes, onions, green pepper, and eggplant. Only 32% of all

participants consumed vitamin A rich fruits and vegetables such as green vegetables,

carrot, mangoes, and papaya (locally called pawpaws). The least consumed food groups

included eggs (15%), dairy/dairy products (25%), and other fruits that are not rich in

vitamin A (45%) (Table 5:3). More than 60% of the sample consumed at least legumes

and nuts (62%); meat, poultry, and fish (67%); and fats and oils (65%).

The proportion of participants with a nutrient intake below the recommended daily

allowance varied between nutrients (Table 5:4). There were no nutrients, for which

participants had sufficient nutrient intake, i.e., for which the nutrient adequate ratio

(NAR) was 100% for all the participants. Nutrients such as carbohydrate and vitamin C

had median NAR above 100%; however, a substantial proportion of participants had

intake below the recommended allowance. Nutrients for which nutrient intake for

participants were 80% or more below the recommended daily allowance included vitamin

D, calcium, iron, and percent protein of energy (Table 5:4).

There were no significant differences in 24-hour dietary nutrient intake deficiencies between men and women for most nutrients except total fat, iron, and folate among HIV

negative individuals without TB; women had a greater proportion of deficiencies

compared to men (Tables 5:5 and 5:6). Women had a deficiency of 57% versus 13%,

100% versus 69%, and 71% versus 25% for total fat, iron, and folate, respectively

compared to men.

All nutrient adequacy expressed as NAR for the different nutrients correlated positively

with Food Variety Score (FVS) and Dietary Density Score (DDS) among both women

and men except percent carbohydrate NAR. For example, the correlation between energy

intake NAR and FVS was 0.69 and it was 0.56 with DDS among women (Table 5:7). All

correlations were significant except percent protein of energy NAR with FVS and DDS

regardless of gender, percent fat of energy NAR with FVS among women and men,

percent carbohydrate of energy with FVS and DDS among women, calcium and vitamin

D NARs with FVS among men. There was also strong positive correlation between FVS

or DDS and the Mean Adequacy Ratio (MAR), the overall score for nutritional adequacy

of the diet (Table 5:7).

The mean MAR value was 63 ± 23 among women and 71 ± 18 among men (Table 5:7).

The ideal cut-off for nutrient adequacy should be 100. This would mean that all nutrients were consumed in adequate amounts. In the present study, none of the participants attained this level. Among women, 59% had MAR greater than or equal to 60; 47% of the women participants had MAR greater than or equal to 65; and 39% had MAR greater

than or equal to 75. Whereas among men, 74% had MAR greater than or equal to 60;

59% of the men participants had MAR greater than or equal to 65; and 44% had MAR

greater than or equal to 75.

We tested different cut-off points of FVS for sensitivity and specificity against different

definitions of a nutritionally adequate diet ranging from MAR = 60 to MAR = 75 as

shown in Figures 5:1 and 5:2 receiver operator curves. We performed this in order to find

an optimal cut-off point for FVS that can identify the maximum inadequate diets as inadequate (high sensitivity), without losing too much ability to identify those with a nutritionally adequate diet (specificity). The figures show that among women and men the sensitivity and specificity were not much influenced by changing the cut-off points of

MAR. We therefore used MAR = 65 as cut-off point for a nutritionally adequate diet. If we were to get a sensitivity of FVS higher than or equal to 75% among women, the cut- off for FVS must be 9 or higher, while to get a specificity of at least 25%, the cut-off for

FVS needs to be 11 or lower. Whereas among men, the cut-off for FVS must be 9 or

higher to get a sensitivity higher than or equal to 75%, while to get a specificity of at least

25%, the cut-off for FVS needs also to be 11 or lower. With a cut-off for FVS on 9 among women and on 9 among men, the sensitivity will be 76% for women and 76% for men and the specificity 70% for women and 67% for men, respectively. If the cut-off for

FVS had been increased by one, the sensitivity would have been increased to 92% among women and compared to 80% among men, but the specificity would decrease to 45% among women and to 50% among men. On the other hand, if the cut-off had decreased by one, the sensitivity would have dropped to 73% among women and 48% among men, and the specificity would have increased to 88% among women and to 83% among men.

We undertook the same exercise to decide the cut-off point for DDS as shown in Figures

5:3 and 5:4. A cut-off point for DDS among women of 5 gave a sensitivity of 70% and a

specificity of 85%. Whereas among men, a cut-off point of 5 gave a sensitivity of 64% and a specificity of 83%. If the cut-off point had been increased to 6 among women, the sensitivity would have been 95% and the specificity of 48% whereas among men the sensitivity would have been 96% and specificity 39%. Decreasing the cut-off point to 4 among women, the sensitivity would have decreased to 32% and the specificity increased to 97% whereas among men the sensitivity would have decreased to 16% and specificity increased to 100%. We therefore took a cut-off point for DDS to be 5 in the present study as the only one point that gives both sensitivity higher than 64% and specificity higher than 80% among both women and men.

Our data shows an increasing MAR with increasing DDS and FVS (Table 5:8). The FVS needs to be 7 to 9 to give a MAR above 65 whereas DDS needs to be at least 5 and above to give satisfactory MAR for women and men. The theoretical estimates of linear regression appear to confirm the data (Table 5:9). To achieve a MAR greater than or equal to 65 with a DDS of 5, one will need FVS of at least 9 among women and 5 among men. For a DDS of 7, the FVS would be 7 among women and as few as 3 among men.

The results from the regression model (Table 5:9, foot note) show that it is only FVS that contributes significantly to the fit of the model among women whereas DDS did fit among men even after adjusting for age, HIV status,

Discussion

The present cross-sectional study that analyzed 131 HIV positive and HIV negative

adults with or without tuberculosis was conducted to assess nutritional adequacy of

dietary intake and validity of simple low-cost methods to evaluate nutritional adequacy of diets consumed among adults in urban Uganda. The dietary consumption in this study population was of low variety and diversity regardless of gender; composed of mostly cereals, roots, tubers, and vegetables not rich in vitamin A. The dietary consumption was

deficient in most nutrients except carbohydrates, vitamin C and B6 that had more than

75% of the study population with 1 nutrient adequate ratio. The deficiency was not

affected by gender, tuberculosis, and HIV status except for iron and folate that differed

by gender among HIV negative individuals. The FVS and DDS increase were associated

with positive increase for nearly all nutrient intake adequacy ratios. The ability of FVS

and DDS indices to identify individuals with inadequate or adequate diet consumption

differed by gender. Prediction of nutritional adequacy differed by gender. The simple

counting of food items (FVS) was a better predictor of nutritional adequacy of the diet

among women whereas counting food groups (DDS) was among men. Thus, FVS and

DDS can provide a rapid and efficient means to health workers to quickly estimate

nutritional adequacy of the diet and also to monitor consumption during management of

tuberculosis and HIV.

Although the interpretation of findings in the present study may be limited by the cross-

sectional nature of the design, specifically use of a single 24-hour dietary recall that the

nutrition intakes do not reflect health endpoints (Vucic et al. 2009) and overall dietary habits (Biro et al. 2002), we employed local utensils with known size and food photographs to enhance recall and estimation of amount for the intake. Further, the dietary recall overlapped with one of the weekend.

Findings in the present study suggest that dietary consumption in this study population was monotonous, rich in carbohydrates and deficient in nutrients. All participants (100%) consumed at least cereals, roots, and tubers, and 90% consumed vegetables not rich in vitamin A such tomatoes and onions while only 45% consumed vitamin-A-rich fruits and vegetables, and only 15% consumed eggs. The mean FVS and DDS for the study population were only 8.1 ± 2.8 and 4.7 ± 1.4, respectively. Both men and women regardless of tuberculosis and HIV status, had carbohydrate and ascorbic acid deficiency in the range of 0 to 30% whereas other nutrient intakes including energy, protein, dietary fiber, calcium, magnesium, zinc, iron, vitamin A, vitamin D, and folate had deficiencies ranging 25% to 100%. Of note, vitamin D had the worst deficiency range from 90% to

100%. This substantial deficiency in vitamin D may a risk to reactivation of tuberculosis

(Sita-Lumbsden A et al. 2007). It is probably difficult to compare results in the present study with other countries on the basis of FVS and DDS because of the different ways in which these indicators have been defined and calculated in different countries (Hatloy,

Torheim, and Oshaug 1998). Further, studies are still limited in populations from sub-

Saharan Africa. Nevertheless, our findings corroborate with the cross-sectional study among the elderly that was conducted in South Africa (Oldewage-Theron and Kruger

2008). Findings in this study revealed low FVS and DDS, and a dietary consumption that

was rich in carbohydrate and nutrient. The strength of our study; however, was the presence of full panel of control groups that comprised of HIV positive and HIV negative adults with or without tuberculosis.

The monotonous diet in this study population reflects on the subsistence economy in

Uganda. Although this study was conducted in urban setting, it appears people purchase mostly agricultural food stuffs similar to the staple food that is grown in their rural home steady. Further, due to lack of public health nutrition education in the general population, people do not know the vital importance of food variety and diversity. People who live at subsistence levels often have no choice but to consume monotonous diets that are poor in nutrients, resulting in poor diet quality (Cannon G 2001). Further, the possible poverty level in this population expose to the monotonous diet. Poor populations suffer most in achieving dietary diversity because of they consume a usual diet based on starchy staple foods containing little or no animal or diary products and few fruit and vegetables, resulting in multiple nutrient deficiencies (Ruel 2003).

Our results show that the FVS and DDS can identify fairly adults with an inadequate nutrient intake. With cut-off points for FVS at 9 and DDS at 5 among women, the indices have high ability to identify those with a nutritionally inadequate and adequate diet whereas among men the indices have lower ability to identify those with inadequate diet, but high ability to identify those with adequate diet. To our knowledge, this is the first study to show gender differences in the ability of FVS and DDS indices to identify

100

individuals or populations with adequate and inadequate diet. Although the value of FVS

and DDS have been established in children (Hatloy, Torheim, and Oshaug 1998; Steyn et

al. 2006; Ruel 2003), limited research has been done in other population groups and disease conditions to explore the consequences of low dietary variety on nutritional status and response during treatment, for example in tuberculosis. This study attempted to define dietary variety across a heterogeneous population with or without tuberculosis and

HIV.

For nutrition promotion and prevention of malnutrition in the population of HIV positive and HIV negative with or without tuberculosis, a high sensitivity is desirable to identify accurately most subjects at nutritional risk (Habicht, Meyers, and Brownie 1982), as long as false positives do not cause other risks. False positives do not pose any risk in the case of the present because of the high prevalence of nutrient deficiencies. In assessment of how select optimal cut-off points for FVS and DDS for an adequate diet, the choice of cut-off point for MAR did not have a strong influence. In our analysis, changing the cut-

off point for MAR from 60 to 75 did not have any vital importance for the conclusions.

We used the liberal 65% as the cut-off point for MAR as used in prior studies (Oldewage-

Theron and Kruger 2008; Steyn et al. 2006). This is supported by the results of Tables 6

and 7, where it is shown that to reach a MAR of 65, one needs FVS and DDS of at least 9

and 5 or higher, respectively.

101

The gender differences in the ability of FVS and DDS indices to identify individuals or populations with adequate and inadequate diet could be explained by the cultural factors that may compromise intake among women. For example, unequal distribution of food within households (Carloni 1981; de Hartog A.P 1972), or men may have the opportunity to eat a wider variety or better quality foods outside the home, such as cafes or local restaurants (Holmboe-Ottesen G and Wandel M 1991). The unequal distribution of food within households result from several factors such as women may be trained to show restraint in eating, to give the best foods to men, or to allow others in the family to eat first (Lado 1992; O'Laughlin B 1974; Rosenberg E.M 1980; Dey J 1981).

The present study revealed that diet consumption among HIV positive and HIV negative adults with or without tuberculosis was monotonous, rich in carbohydrates and deficient in nutrients. The ability of FVS and DDS indices to identify individuals with inadequate or adequate diet consumption differs by gender. The simple counting of food items (FVS) was a better predictor of nutritional adequacy of the diet among women whereas counting food groups (DDS) was among men. Thus, FVS and DDS can provide a rapid and efficient means to health workers to quickly estimate nutritional adequacy of the diet, to monitor dietary intake in management of tuberculosis and HIV, and to assess patient’s nutritional knowledge and thereafter provide health education. Validation studies are; however, needed in follow-up studies and other regions of the country.

102

Acknowledgements

We thank all study staff members of the Case Western Reserve University and Makerere

University research collaboration at the Tuberculosis Research Unit in the United States and in Uganda for their assistance; the faculty of staff at Case Western Reserve

University Department of Epidemiology and Biostatistics for the guidance in analyzing the project; and the Fogarty International Center, for the continued support.

This study was supported in part by the AIDS International Training Research Program,

Fogarty International Center, Grant No. TW000011, based at Case Western Reserve

University, Department of Epidemiology and Biostatistics. This work was part of Ezekiel

Mupere’s PhD thesis at Case Western Reserve University.

103

Table 5:1 Characteristics of adult participants with tuberculosis in Kampala,

Uganda

Characteristic HIV positive with TB HIV negative with TB

[mean, (SD)] (n=31) (n=32)

Women Men Women Men

(n=21) (n=10) (n=14) (18)

Age in yrs 29.2 (5.9) 30.9 (4.6) 26.0 (5.4) 26.0 (7.3)

Weight in kg 46.8 (7.3) 54.3 (6.1)b 50.0 (7.8) 53.2 (6.5)

Height in cm 158.9 (6.7) 171.8 (9.3)a 157.4 (13.2) 171.1 (6.0)a

BMI in kg/m2 18.6 (2.9) 18.4 (1.7) 20.5 (4.4) 18.2 (2.0)

FVS 7.1 (2.8) 6.4 (1.7) 8.7 (2.8) 8.7 (2.7)

DDS 4.5 (1.7) 4.7 (1.5) 5.4 (1.6) 4.7 (1.1)

MAR 61.4 (24.3) 68.0 (17.6) 68.7 (17.3) 70.6 (17.4)

ap-value <0.004, bp-value <0.05. FVS = Food Variety Score: the number of all food items used in the study period (1 – 127). DDS = Dietary Diversity Score: number of food groups eaten (1 – 8): cereals, roots, and tubers; vitamin A-rich fruits and vegetables; other fruits not rich in vitamin A; other vegetables not rich in vitamin A; legumes and nuts; meat, poultry, and fish; fats and oils; diary and products; and eggs. MAR = Mean Adequacy Ratio: ratio of 11 micronutrients: vitamin A, vitamin B6, vitamin C, vitamin D, thiamin, riboflavin, folate, magnesium, calcium, iron, and zinc.

104

Table 5:2 Characteristics of adult participants without tuberculosis in Kampala,

Uganda

Characteristic HIV positive with no TB HIV negative with no TB

[mean, (SD)] (n=38) (n=30)

Women Men Women Men

(n=21) (17) (n=14) (16)

Age in yrs 29.7 (8.4) 34.7 (6.5)b 24.3 (5.4) 22.4 (3.2)

Weight in kg 57.7 (9.4) 62.1 (6.6) 60.7 (8.7) 59.8 (5.6)

Height in cm 155.0 (5.6) 170.5 (7.9)a 159.7 (6.0) 167.2 (6.8)b

BMI in kg/m2 24.2 (4.6) 21.4 (2.3)b 23.7 (2.8) 21.4 (2.3)b

FVS 8.2 (3.4) 8.5 (2.4) 7.6 (2.5) 9.6 (2.6)

DDS 4.3 (1.4) 4.8 (1.0) 4.5 (1.5) 5.1 (1.1)b

MAR 60.2 (27.3) 68.5 (17.1) 62.6 (19.5) 74.3 (20.0)

105

Table 5:3 Food groups and food items from 24-hour dietary intake recall among

HIV positive and HIV negative adults in Kampala, Uganda (n=131)

Food groups Frequency (%) % Food item

Cereals, roots, and 100.0 31.7 Green plantain (Musa acuminata) tubers 10.3 Maize meal (Zea mays)

9.9 Sweet potatoes (Ipomoea batatas)

9.0 Rice (Oryza sativa)

Vitamin-A-rich fruits 31.6 34.2 Green vegetables (Amaranthus) and vegetables 23.7 Carrot (Daucus carota)

26.3 Mango fruit (Mangifera)

13.2 Pawpaw (Papaya, Carica)

Other fruits not rich in 44.5 41.7 Passion fruit (Passiflora edulis) vitamin A 25.0 Ripe banana (Musa sapientum)

9.7 Avocado (Persea Americana)

9.7 Pineaple (Ananas comosus)

Other vegetables not 89.8 43.9 Tomato (Solanum lycopersicum) rich in vitamin A 24.4 Onion (Allium cepa)

106

14.6 Green pepper (Capsicum annum)

2.4 Eggplant (solanum melongena)

Legumes and nuts 62.1 45.4 Groundnuts (Arachis hypogaes)

44.4 Beans (Vigna Angularis)

5.6 Soyabeans (Glycine max)

4.6 Cowpeas (Vigna unguiculata)

Meat, poultry, and fish 67.1 53.3 Beef

35.8 Fish

7.5 Chicken

Fats and oils 64.9 53.9 Ghee

23.1 Cooking oil

23.1 Margarine (Blue-band)

Dairy 24.5 76.9 Milk with tea

23.1 Whole milk

Eggs 15.1 84.2 Boiled eggs

These food groups were the basis of Dietary Diversity Score. In Food Variety Score all food items were included.

107

Table 5:4 The 24-hour dietary intake recall, recommended daily allowance, frequency of inadequate intake of nutrients among HIV positive and HIV negative adults in Kampala, Uganda (n=131)

24-Hour Intake RDA Deficit Median Nutrient Median Q25 Q75 Women Men (NAR) intake (%)

Energy, kcal 1732 1093 2345 2700 2000 75.5 73

Protein, g 40.5 29.0 62.1 56 46 83.8 60

Total fat, g 36.8 22.2 59.0 30 30 122.7 39

Carbohydrate, g 287 188 414 130 130 144.2 9

Dietary fiber, g 30.8 18.2 49.7 38 25 97.2 53

Protein, % energy 10.0 8.0 12.0 12.5 12.5 80.0 80

Fat, % energy 19.0 14.0 26.0 22.5 22.5 84.4 39

CHO, % energy 70.0 62.0 75.0 60 60 116.7 9

Calcium, mg 196 101 499 1000 1000 19.6 95

Magnesium, mg 177 104 315 260 220 68.9 64

Zinc, mg 5.2 3.1 7.6 7.0 4.9 92.9 56

Iron, mg 7.9 4.9 13.0 14 29 38.6 86

108

Vitamin A, μg 208 70 551 600 500 35.7 74

Thiamin, mg 0.8 0.5 1.2 1.2 1.1 66.7 70

Riboflavin, mg 0.9 0.5 1.3 1.3 1.1 72.7 66

Vitamin B6 2.2 1.3 3.1 1.3 1.3 169.2 22

Vitamin C, mg 91 49 178 45 45 202.8 22

Vitamin D, μg 0 0 1.0 5 5 0 98

Folate, μg 341 172 473 400 400 85.3 63

RDA = Recommended Daily Allowance, CHO = carbohydrate. NAR (Nutrient Adequacy Ratio) = actual intake/recommended intake. FAO = Food and Agriculture Organization, WHO = World Health Organization 2002. All NAR for micronutrients estimated using FAO/WHO requirements 2002 except vitamin D which was based on adequate intake (Atkinson and Ward 2001). DRFI = Dietary Reference Intake. NAR for energy, protein, carbohydrates, fat, and dietary fiber were estimated using DRFI.

109

Table 5:5 Percent of inadequate 24-hour dietary recall intake among HIV positive and HIV negative adults with tuberculosis in Kampala, Uganda

Characteristic HIV+ with TB (n=31) HIV- with TB (n=32)

Men Women Men Women

(n=10) (n=21) (n=18) (n=14)

Energy (kcal) n (%) 8 (80) 16 (76) 12 (67) 10 (71)

Protein (g) n (%) 6 (60) 14 (67) 10 (56) 10 (71)

Total fat (g) n (%) 3 (30) 14 (67) 5 (28) 6 (43)

Carbohydrate (g) n (%) 1 (10) 6 (29) 0 (0) 1 (7)

Dietary fiber (g) n (%) 6 (60) 11 (52) 9 (50) 6 (43)

Protein, % energy n (%) 8 (80) 13 (62) 16 (89) 10 (71)

Fat, % energy n (%) 4 (40) 16 (76) 10 (56) 9 (64)

Carbohydrate, % energy n (%) 4 (40) 4 (19) 5 (28) 3 (21)

Calcium (mg) n (%) 8 (80) 20 (95) 17 (94) 14 (100)

Magnesium (mg) n (%) 9 (90) 13 (62) 8 (44) 10 (71)

Zinc (mg) n (%) 6 (60) 11 (52) 10 (56) 9 (64)

Iron (mg) n (%) 8 (80) 19 (90) 14 (78) 14 (100)

110

Vitamin A (RE) n (%) 7 (70) 15 (71) 14 (82) 6 (43)

Vitamin D (μg) n (%) 9 (90) 20 (95) 18 (100) 14 (100)

Ascorbic acid (mg) n (%) 2 (20) 6 (30) 5 (28) 2 (14)

Folate (μg) n (%) 8 (80) 15 (71) 13 (72) 7 (50) ap-value <0.001, bp-value <0.05. HIV+ = HIV positive, HIV- = HIV negative, TB = tuberculosis.

111

Table 5:6 Percent of inadequate 24-hour dietary recall intake among HIV positive and HIV negative adults without tuberculosis in Kampala, Uganda

HIV- with no TB HIV+ with no TB Characteristic (n=38) (n=30)

Women Men Men Women

(n=17) (21) (n=16) (n=14)

Energy (kcal) n (%) 15 (88) 14 (67) 11 (69) 10 (71)

Protein (g) n (%) 8 (47) 14 (67) 6 (38) 10 (71)

Total fat (g) n (%) 5 (29) 8 (38) 2 (13) 8 (57)b

Carbohydrate (g) n (%) 0 (0) 4 (19) 0 (0) 0 (0)

Dietary fiber (g) n (%) 8 (47) 13 (62) 7 (44) 9 (64)

Protein (% of energy) n (%) 13 (76) 17 (81) 15 (94) 13 (93)

Fat (% of energy) n (%) 10 (59) 14 (67) 9 (56) 8 (57)

Carbohydrate (% of energy) n (%) 3 (18) 2 (10) 0 (0) 4 (29)b

Calcium (mg) n (%) 16 (94) 20 (95) 16 (100) 13 (93)

Magnesium (mg) n (%) 13 (76) 14 (67) 8 (50) 9 (64)

Zinc (mg) n (%) 10 (59) 11 (52) 7 (44) 10 (71)

112

Iron (mg) n (%) 14 (82) 19 (90) 11 (69) 14 (100)b

Vitamin A (RE) n (%) 15 (88) 12 (60) 14 (88) 12 (86)

Vitamin D (μg) n (%) 17 (100) 21 (100) 14 (100) 16 (100)

Ascorbic acid (mg) n (%) 4 (23) 3 (15) 3 (19) 2 (14)

Folate (μg) n (%) 11 (65) 14 (67) 4 (25) 10 (71)b ap-value <0.001, bp-value <0.05. HIV+ = HIV positive, HIV- = HIV negative, TB = tuberculosis.

113

Women Men

(n=70) (n=61)

NAR Nutrient Food Dietary Food Dietary

Variety Diversity Variety Diversity

Score1 Score2 Score1 Score2

NAR Energy (kcal) 0.69 0.56 0.53 0.54

NAR Protein (g) 0.66 0.53 0.50 0.54

NAR Total fat (g) 0.59 0.53 0.54 0.57

NAR Carbohydrate (g) 0.59 0.49 0.46 0.38b

NAR Dietary fiber (g) 0.50 0.52 0.26b 0.37b

NAR Protein (% of energy) 0.18* 0.18* 0.11* 0.10*

NAR Fat (% of energy) 0.18* 0.21b 0.23* 0.26b

NAR Carbohydrate (% of energy) -0.22* -0.20* -0.28b -0.29b

NAR Calcium (mg) 0.70 0.69 0.22* 0.39

NAR Magnesium (mg) 0.63 0.51 0.43 0.31b

114

NAR Zinc (mg) 0.59 0.44 0.39b 0.34b

NAR Iron (mg) 0.59 0.49 0.29b 0.38b

NAR Vitamin A (RE) 0.50 0.59 0.30b 0.61

NAR Thiamin (mg) 0.59 0.51 0.27b 0.39b

NAR Riboflavin (mg) 0.61 0.55 0.35b 0.60

b NAR Vitamin B6 0.48 0.46 0.40 0.49

NAR Ascorbic acid (mg) 0.38b 0.42 0.36b 0.45

NAR Vitamin D (μg) 0.29b 0.44 0.08* 0.35b

NAR Folate (μg) 0.43 0.43 0.62 0.52

Mean Adequacy Ratio (MAR) 0.70 0.64 0.45 0.60

All correlation p-values <0.001 except bp-value < 0.05, *p-value >0.05. 1Number of all food items used in the study period. 2Number of all food groups consumed (1 – 9): 1) Cereal, roots, and tubers, 2) vitamin-A-rich fruits and vegetables, 3) 0ther fruits not rich in vitamin A, 4) other vegetables not rich in vitamin A, 5) legumes and nuts, 6) meat, poultry, and fish, 7) fats and oils, 8) diary, and 9) eggs and eggs products.

115

Table 5:8 Mean MAR scores for different levels of Food Variety Score (FVS) and

Dietary Diversity Score (DDS) among HIV positive and HIV negative adults in

Kampala, Uganda (n=131)

Dietary Food Variety Score

Diversity Score 1-3 4-6 7-9 10-12 13-15

Women (n=70)

1 24 (2) - - - -

2 26 (n=2) 46 (n=5) - - -

3 - 59 (n=4) - - -

4 - 39 (n=8) 65 (n=9) 63 (n=1) -

5 - 44 (n=3) 65 (n=12) 71 (n=5) 97 (n=1)

6 - - 73 (n=4) 87 (n=4) 90 (n=2)

7 - - 77 (n=3) 93 (n=2) 79 (n=1)

8 - - - 85 (n=1) 90 (n=1)

9 - - - - -

Men (n=61)

1 37 (n=2) - - - -

116

2 - 57 (n=1) - - -

3 - 65 (n=1) - - -

4 - 58 (n=7) 58 (n=9) 68 (n=2) -

5 - 73 (n=4) 73 (n=7) 75 (n=12) 90 (n=1)

6 - - 73 (n=5) 89 (n=7) -

7 - - 91 (n=1) 80 (n=1) -

8 - - - 92 (n=1) -

9 - - - - -

117

Table 5:9 Estimated Mean Adequacy Ratio scores for different levels of Food

Variety Score (FVS) and Dietary Diversity Score (DDS) from Linear Regression

Model among HIV positive and HIV negative adults in Kampala, Uganda (n=131)

Dietary Food Variety Score

Diversity

Score 3 5 6 7 8 9 10 12 15 (DDS)

Women (n=70)

1 32 40 44 49 53 57 61 70 82

2 35 43 47 51 56 60 64 73 85

3 38 46 50 54 59 63 67 75 88

4 41 49 53 57 62 66 70 78 91

5 43 52 56 60 64 69 73 81 94

6 46 55 59 63 67 72 76 84 98

7 49 58 62 66 70 74 79 87 100

8 52 61 65 69 73 77 82 90 103

9 55 63 68 72 76 80 84 93 105

Men (n=61)

118

1 33 36 37 39 40 41 43 46 50

2 41 43 45 46 48 49 51 54 58

3 48 51 53 54 56 57 58 61 66

4 56 59 60 62 63 65 66 69 74

5 64 67 68 70 71 73 74 77 81

6 72 75 76 77 79 80 82 85 89

7 82 82 84 85 87 88 90 93 97

8 87 90 92 93 94 96 97 100 105

9 95 98 99 101 102 104 105 108 112

Regression model for women: MAR = 16.349 + 2.906*DDS + 4.196*FVS. For men: MAR = 20.593 + 7.785*DDS + 1.453*FVS

119

Linear Regression Model to Estimate MAR Scores for Women

Dependent variable: MAR; Predictors: DDS and FVS

R2 Adjusted R2 Coefficient of variation p-value

0.502 0.487 26.138 <0.001

Unstandardized coefficients

β SE T p-value

Constant 16.349 6.211 2.63 0.011

DDS 2.906 2.071 1.40 0.165

FVS 4.196 1.106 3.80 0.0003

120

Linear Regression Model to Estimate MAR Scores for Men

Dependent variable: MAR; Predictors: DDS and FVS

R2 Adjusted R2 Coefficient of variation p-value

0.408 0.387 19.752 <0.001

Unstandardized coefficients

β SE T p-value

Constant 20.593 8.118 2.54 0.014

DDS 7.785 1.932 4.03 0.0002

FVS 1.453 0.834 1.74 0.870

121

Figure 5:1 Sensitivity (sens) and specificity (spec) for different cut-off points of Food

Variety Score (FVS) among women: Mean Adequacy Ratio (MAR) changing from

60 to 75

120

100

MARSe: 60 80 MARSe: 65 MARSe: 70 60 MARSe: 75 MARSp: 60 40 MARSp: 65 MARSp: 70 % % sensitivityand specificity 20 MARSp: 75

Cut-off point for FVS 0 1' 2' 3' 4' 5' 6' 7' 8' 9' 10' 11' 12' 13' 14' 15' MARSe: 60 0 7 7 14 31 48 69 76 79 97 100 100 100 100 100 MARSe: 65 0 5 5 11 24 41 62 73 76 92 97 100 100 100 100 MARSe: 70 0 5 5 10 22 37 56 66 76 93 98 100 100 100 100 MARSe: 75 0 5 5 9 21 35 53 65 77 93 98 100 100 100 100 MARSp: 60 100 100 100 100 100 98 90 78 63 41 29 17 12 5 2 MARSp: 65 100 100 100 100 100 100 97 88 70 45 33 21 15 6 3 MARSp: 70 100 100 100 100 100 100 97 86 76 52 38 24 17 7 3 MARSp: 75 100 100 100 100 100 100 96 89 81 56 41 26 19 7 4

Sensitivity MAR 60 – 75 = Sensitivity for a given cut-off point of FVS with a cut-off point for MAR varying from 60 to 75. Specificity MAR 60 – 75 = Sensitivity for a given cut-off point of FVS with a cut-off for MAR varying from 60 to 75. Sensitivity = identify nutritionally inadequate diets as inadequate; specificity = identify nutritionally adequate diets as adequate.

122

Figure 5:2 Sensitivity (sens) and specificity (spec) for different cut-off points of Food

Variety Score (FVS) among men: Mean Adequacy Ratio (MAR) changing from 60

to 75

120

100

MARSe: 60 80 MARSe: 65 MARSe: 70 60 MARSe: 75 MARSp: 60 40 MARSp: 65 MARSp: 70 % % sensitivityand specificity 20 MARSp: 75

Cut-off point 0 for FVS 1' 2' 3' 4' 5' 6' 7' 8' 9' 10' 11' 12' 13' 14' 15' MARSe: 60 0 0 6 13 19 31 38 44 69 75 88 88 100 100 100 MARSe: 65 0 0 4 8 12 28 44 48 76 80 88 92 100 100 100 MARSe: 70 0 0 3 7 13 27 40 43 70 83 90 93 100 100 100 MARSe: 75 0 0 3 6 15 26 38 41 68 82 88 94 100 100 100 MARSp: 60 100 100 100 100 96 91 80 76 56 42 31 16 2 0 0 MARSp: 65 100 100 100 100 94 94 89 83 67 50 36 19 3 0 0 MARSp: 70 100 100 100 100 97 97 90 84 68 58 42 23 3 0 0 MARSp: 75 100 100 100 100 100 100 93 85 70 63 44 26 4 0 0

123

Figure 5:3 Sensitivity (sens) and specificity (spec) % for different cut-off points of

Diet Diversity Score (DDS) among women: Mean Adequacy Ratio (MAR) changing

from 60 to 75

120

100 MARSe: 60 80 MARSe: 65 MARSe: 70

60 MARSe: 75 MARSp: 60 MARSp: 65 % % sensitivityand specificity 40 MARSp: 70 MARSp: 75 20

Cut-off points for DDS 0 1' 2' 3' 4' 5' 6' 7' 8' 9' MARSe: 60 0 7 28 38 72 100 100 100 100 MARSe: 65 0 5 24 32 70 95 100 100 100 MARSe: 70 0 5 22 29 63 93 98 100 100 MARSe: 75 0 5 21 28 63 93 98 100 100 MARSp: 60 100 100 98 95 76 44 20 5 0 MARSp: 65 100 100 100 97 85 48 24 6 0 MARSp: 70 100 100 100 97 83 52 24 7 0 MARSp: 75 100 100 100 96 85 56 26 7 0

Sensitivity MAR 60 – 75 = Sensitivity for a given cut-off point of DDS with a cut-off point for MAR varying from 60 to 75. Specificity MAR 60 – 75 = Sensitivity for a given cut-off point of DDS with a cut-off for MAR varying from 60 to 75. Sensitivity = identify nutritionally inadequate diets as inadequate; specificity = identify nutritionally adequate diets as adequate.

124

Figure 5:4 Sensitivity (sens) and specificity (spec) % for different cut-off points of

Diet Diversity Score (DDS) among men: Mean Adequacy Ratio (MAR) changing

from 60 to 75

120

100 MARSe: 60 80 MARSe: 65 MARSe: 70

60 MARSe: 75 MARSp: 60 MARSp: 65 % % sensitivityand specificity 40 MARSp: 70 MARSp: 75 20

Cut-off points for DDS 0 1' 2' 3' 4' 5' 6' 7' 8' 9' MARSe: 60 0 0 19 19 69 100 100 100 100 MARSe: 65 0 0 12 16 64 96 100 100 100 MARSe: 70 0 0 10 13 63 97 100 100 100 MARSe: 75 0 0 9 12 59 91 100 100 100 MARSp: 60 100 100 100 98 76 33 7 2 0 MARSp: 65 100 100 100 100 83 39 8 3 0 MARSp: 70 100 100 100 100 90 45 10 3 0 MARSp: 75 100 100 100 100 93 44 11 4 0

125

CHAPTER 6

PREDICTORS OF FAT MASS AND LEAN TISSUE AMONG HIV POSITIVE

AND HIV NEGATIVE ADULTS WITH OR WITHOUT TUBERCULOSIS IN

URBAN KAMPALA, UGANDA

126

Abstract

Background Fat and fat-free mass body composition measurements have been reported

to permit a more precise evaluation of body wasting and malnutrition than use of body

mass index; however, factors that may influence fat and fat-free mass and body mass index (BMI) have not been well characterized.

Objective We evaluated whether energy and protein intake are predictors of fat and fat- free mass and BMI in a population of HIV positive and HIV negative adults with or without tuberculosis in urban Kampala, Uganda.

Methods In a cross-sectional study with 131 adults who were screened for active tuberculosis and HIV infection, energy and protein intakes using a 24-hour dietary recall, weight, height, fat and fat-free mass using bioelectrical impedance analysis were assessed.

Results Energy intake was associated with an increase in BMI among women although intake in the presence of tuberculosis was associated with a decrease in BMI. Protein intake among women with no income and among women with unemployment was associated with a decrease in fat-free mass and fat mass, respectively whereas protein intake among women with tuberculosis was associated with an increase in BMI. Being single among women was associated with an increase in fat-free mass whereas having reduced appetite was associated with a decrease in fat-free mass and fat mass. Among men, tuberculosis was associated with a decrease in fat-free mass. Similarly, having reduced appetite was associated with decrease in fat mass and BMI. HIV did not influence body composition regardless of gender.

127

Conclusion The present study has revealed that there are remarkable gender differences in how energy and protein intakes influence body composition, and there are important interactions with presence of tuberculosis, lack of income, and unemployment. HIV does not appear to influence nutrient intake on body composition. Further evaluation is needed to understand how these differences would influence body composition over time and survival.

128

Background

Tuberculosis and human immunodeficiency virus (HIV) are diseases of major public

health concern worldwide, with the highest burden in sub-Saharan Africa (WHO 2009;

Lawn and Churchyard 2009). Among people with tuberculosis and HIV co-infection,

tuberculosis is the leading cause of death in sub-Saharan Africa (Lawn and Churchyard

2009; Corbett et al. 2003).

Both tuberculosis and HIV are independently associated with body wasting, and the

wasting in tuberculosis patients is thought to be further exacerbated by the concomitant

effects of HIV (Kotler 2000; Macallan 1999; Lucas et al. 1994). In contrast, findings from several cross-sectional studies (Niyongabo et al. 1999; Mupere et al. 2010; Paton and Ng 2006) appear to show no significant differences in body composition between

HIV positive adults with tuberculosis and HIV negative adults with tuberculosis suggesting that tuberculosis is the primary factor in driving the process wasting during co-infection. However, gender differences in body composition at presentation among tuberculosis patients have been reported (Kennedy et al. 1996; Mupere et al. 2010).

Although poorly understood, the wasting process in tuberculosis and HIV is probably due to a number of factors, including reduced nutrient intake, malabsorption, altered metabolism, and interactions between drugs and nutrients (Paton et al. 1999; Paton et al.

2003; Macallan et al. 1998; Fields-Gardner 1995; Kotler and Grunfeld 1995).

129

Fat and fat-free mass body composition measurements have been reported to permit a

more precise evaluation of body wasting than use of body mass index (BMI) (Kyle,

Piccoli, and Pichard 2003; VanItallie et al. 1990; Kyle et al. 2003). Evaluation of fat and

fat-free mass allows specific understanding of the body compartment and the extent to which the compartment is involved in the wasting. Moreover, disproportionate loss of fat-

free mass referred to as lean tissue may be associated with adverse effects on survival and

physical function.

BMI is insensitive to body fatness, particularly at low BMI, as well as with above-normal

muscle development (Kyle, Genton, and Pichard 2002; Kyle, Piccoli, and Pichard 2003).

The use of height-normalized fat and fat-free mass indexes (fat mass index (FMI) and fat-

free mass index (FFMI), respectively) allows partitioning of BMI into fat mass index

(FMI) and fat-free mass index (FFMI), i.e., BMI = FFMI + FMI (VanItallie et al. 1990;

Kyle et al. 2003). The height-normalized FMI and FFMI eliminate differences in fat and

fat-free mass associated with height (Baumgartner et al. 1998). Bioelectrical impedance

analysis (BIA) has been recommended as the preferred and precise method for clinical

assessment of FFM and fat mass (Kyle, Genton, and Pichard 2002; Kyle et al. 2004).

Prior studies conducted in African patients (Niyongabo et al. 1999; Villamor et al. 2006)

using BIA have reported reduced fat and fat-free mass at the time when patients present

with tuberculosis. However, the factors that may influence fat and fat-free mass indexes and BMI have not been well characterized.

130

In this cross-sectional study, we evaluated whether energy and protein intake are

predictors of fat and fat-free mass indexes and BMI in a population of HIV positive and

HIV negative with or without tuberculosis in urban Kampala, Uganda.

Methods

We enrolled 132 participants in a cross-sectional study with age 18 or more years residing in Kampala district or 20 km from the study site if residence was outside

Kampala in Uganda. Data collection was conducted between November 2007 and March

2008, a period that coincides with harvesting and light rains in November and December and dry season in January and February. One participant was excluded from the analysis because of prior tuberculosis treatment. The study was conducted at the National

Tuberculosis and Leprosy Program (NTLP) Clinic of the national tertiary teaching hospital, Mulago complex. Of the 131 participants who were included in the analysis, 31 were HIV positive and 32 HIV negative tuberculosis patients who were recruited at the

Mulago NTLP Clinic; 38 were HIV positive without tuberculosis and recruited at the

Infectious Disease Institute Clinic (IDI) located 500 meters from the Mulago NTLP

Clinic; and 30 were HIV negative adults without tuberculosis from the community where enrolled tuberculosis patients resided. The institutional review boards at Case Western

Reserve University and Joint Clinical Research Center approved the study, with final approval by the Uganda National Council for Science and Technology. All participants provided written informed consent to the study.

131

All subjects in the study were given appropriate pre- and post-test HIV counseling and

AIDS education. HIV-1 infection was diagnosed on the basis of a positive enzyme-linked immunosorbent assay for HIV-1 antibodies (Recombigen; Cambridge Biotech,

Patients with active tuberculosis were treated with standard four-drug chemotherapy for tuberculosis per guidelines of the Ugandan Ministry of Health. Similarly, HIV positive patients eligible for antiretroviral therapy were started on treatment and cotrimoxazole prophylaxis at the IDI clinic.

Anthropometric measurements included height and weight. Body-mass index (BMI) was computed using the relationship of weight in kilograms divided by height in meters squared (kg/m2). Weight was taken using Hanson digital electronic bathroom weighing

scales to the nearest 100g. Height was measured to the nearest mm by calibrated standing

height using a stadiometer. All anthropometric measurement values were the mean of

duplicates. The single-frequency bioelectrical impedance analyzer (BIA Detroit, MI, RJL

Systems) performing at 50 kHz and 800 mA was used for BIA measures with detecting

electrodes placed on the wrist and ankle and signal introduction electrodes placed on the

first joint of the middle finger and behind the middle toe. Before performing

measurements on each subject, the BIA instrument was calibrated using the

manufacturer’s recalibration device. The resistance and reactance were based on

measures of a series circuit (Kotler et al. 1996). BIA measurements were performed in

132

duplicate for each subject. The analyzer was calibrated monthly. Fat-free mass was

calculated from BIA measurements using equations that were previously cross-validated

in a sample of patients (white, black and Hispanic) with and without HIV infection

(Kotler et al. 1996) and have been applied elsewhere in African studies (Shah et al. 2001;

Van Lettow et al. 2004; Villamor et al. 2006). Fat mass was calculated as body weight

minus fat-free mass.

The dietary intake assessment was made using a single 24-hour dietary recall with open

ended questions. The reference period for the 24-hour recall was the day prior to the day of the interview. In all instances, the interview was held only with the interviewee, no one else was present except for children. The questionnaire was pre-tested by administering it to 8 individuals selected randomly from the neighboring community to the study site. The assessment was conducted by four trained study nursing staffs and continuously supervised by a nutritionist using local food photographs, portion-size images, and

volumetric vessels to increase the accuracy of the recall from the previous 24 hours. The

nutritive value of raw ingredients was computed using the East African food composition

table database whose database was imported into the NutriSurvey software

(http://www.nutrisurvey.de) to easy the computations. The database used was

predominantly for local Ugandan diet. When the East African food composition table was

found deficient in certain food items, the United States Department of Agriculture

database and the African composition table were used.

133

Analytic strategy

Spearman correlations were performed between fat-free mass, fat mass, or BMI and energy or protein intake to establish the strengths of association. To establish whether energy and protein intakes are predictors of fat-free mass, fat mass, and BMI, we performed multivariable linear regression analyses with energy and protein intakes as main independent variables including all variables that were associated with a p < 0.50 in the unadjusted analyses (Dales and Ury 1978). Height-normalized indexes of fat-free mass, fat mass, and body mass were used in the models. The following variables were adjusted for in multivariable analyses: older age group >30 years, having tuberculosis and

HIV infection, no or low level (primary) of education, being single, being separated or divorced, having a household number of more than 2 people, unemployment, having no personal income, current history of alcohol intake, and reduced appetite.

Both univariate and multivariable analyses have been presented in this article. Previous

findings (Mupere et al. 2010) revealed gender differences in body composition among

individuals with or without tuberculosis regardless of HIV status. Further, there was

significant interaction between gender and energy intake. Thus, we stratified our analysis

according to gender. Two-way interactions were tested between energy or between

protein intakes for all variables involved in each multivariable model. The R-square was

used to evaluate the importance variables in the model. Influence diagnostics of the final

results was conducted to ensure that the results were not unduly influenced by a few

outlying measurements (Vittinghoff E. et al. 2005). The final regression coefficients were

134

obtained after eliminating participants with influential observations (Neter J., Wasserman

W., and Kutner M.H. 1990). Removal of influential points yielded more conservative

standard errors of the estimates. All analyses were performed using SAS version 9.2

(Cary software, North Carolina SAS Institute Inc 2004.)

Results

The characteristics of the study population are shown in Table 6:1 and 6:2. Of the 131

participants who were included in the analysis, 53% were females, 47% were males, 53%

were HIV positive, and 48% had tuberculosis.

The strengths of association between energy or protein intake and fat-free mass, fat mass,

or BMI are shown by spearman’s correlation coefficients (Table 6:3). Energy and protein

intakes had positive correlations with fat-free mass, fat mass and BMI regardless of gender. Among women, there were significant correlations between energy or protein intake and fat-free mass, fat mass, BMI, or with height-normalized indexes of fat and fat-

free mass.

Predictors of fat-free mass

In univariate analysis for the total study population, energy and protein intakes and being

single were associated with an increase in fat-free mass whereas tuberculosis, lack of any income, and having reduced appetite were associated with a decrease in fat-free mass

135

(Tables 6:4 and 6:5). Of note, energy intake among women was associated with a double

0.0004 ± 0.0002 standard error (SE) (p=0.028) increase in fat-free mass compared to men

that had 0.0002 ± 0.0003 SE, p=0.418 (Table 6:5). Having tuberculosis and reduced

appetite were associated with a decrease in fat-free mass regardless of gender.

In multivariable analysis for the total study population, there was significant interaction

between energy intake and gender when the model had gender, HIV status, tuberculosis

status, being single, being more two people per household, unemployment status, lack of

any income, and reduced appetite as adjusters. Energy intake among women was

associated with a significant decrease in fat-free mass, -0.001 ± 0.0003 SE, p=0.004). We thus performed stratified models according to gender adjusting for the same variables but without gender (Table 6:6 and 6:7). In the overall multivariable model, energy intake was associated with an increase in fat-free mass; however, this effect was not noticeable after stratification by gender (Table 6:6 and 6:7). Having tuberculosis was associated with a significant decrease in fat-free mass for the overall population; however, this decrease was double among men compared to women. Having reduced appetite was associated with a significant decrease in fat-free mass. There was a significant interaction between protein intake and income. Protein intake among women with no income was associated with a significant decrease in fat-free mass (Table 6:7). Of note, HIV was not associated with prediction of fat-free mass regardless of gender.

136

Predictors of fat mass

Energy intake in univariate analysis was associated with a decrease in fat mass of -

0.00002 ± 0.0003 SE, p=0.945 for the total population; however, this decrease was noted among men. Energy intake among women was associated with a significant increase in fat mass of 0.001 ± 0.0005 SE, p=0.020 (Table 6:8 and 6:9). Having tuberculosis and reduced appetite were associated with a decrease in fat mass regardless of gender. Of note, HIV was not associated with prediction of fat mass.

There was significant interaction between energy intake and gender in multivariable analysis for the overall study population when energy and protein intakes, gender, tuberculosis, being more two people per household, unemployment, lack of any income, and reduced appetite were independent variables in the model (Table 6:10 and 6:11).

Energy intake among women was associated with a significant increase in fat mass, 0.002

± 0.001, SE, p=0.001. Thus, in stratified analysis according to gender adjusting for the same variables, having reduced appetite was associated with a decrease in fat mass regardless of gender. Of note, women had a fourfold decrease (-4.17 ± 1.94, SE) in fat mass compared to men (-1.90 ± 0.90, SE) (Table 6:11). There was a significant interaction between protein intake and unemployment. Lack of employment among women was associated with a decrease in fat mass.

137

Predictors of BMI

In univariate analysis, energy intake was associated with an increase in BMI of 0.002 ±

0.001 SE, p=0.014 among women (Table 6:12 and 6:13). Energy intake among men was not significant. Having tuberculosis and reduced appetite were associated with a decrease in fat mass regardless of gender. Of note, HIV was not associated with prediction of fat mass (Table 6:12 and 6:13).

In multivariate analysis, after adjusting for protein intake, tuberculosis, lack of income, and reduced appetite, energy intake was associated an increase of 0.003 ± 0.001 ± SE among women (Table 6:14 and 6:15). Energy intake among men was not significant; however, having reduced appetite was associated with a four unit decrease in BMI among men. There were significant interactions between energy intake and tuberculosis, between energy intake and appetite, and between protein intake and tuberculosis (Table 6:14 and

6:15). Energy intake in the presence of tuberculosis among women was associated with a significant decrease in BMI whereas protein intake in the presence of tuberculosis was associated with a significant increase in BMI among women. None of these were significant among men. Energy intake with reduced appetite was associated with an increase in BMI in the overall population; however, this interaction not significant among women and among men (Table 6:14 and 6:15).

138

Discussion

In the present cross-sectional study of 131 participants, we evaluated whether energy and

protein intakes were important predictors of fat-free mass, fat mass, and BMI in a population of HIV positive and HIV negative adults with or without tuberculosis in urban

Uganda. Predictors of fat and fat-free mass and BMI differed by gender. Energy intake was associated with an increase in BMI among women although intake in the presence of tuberculosis was associated with a decrease in BMI. Protein intake among women with no income and among women with unemployment was associated with a decrease in fat- free mass and fat mass, respectively whereas protein intake among women with tuberculosis was associated with an increase in BMI. Being single among women was associated with an increase in fat-free mass whereas having reduced appetite was associated with a decrease in fat-free mass and fat mass. Among men, tuberculosis was associated with a decrease in fat-free mass. Similarly, having reduced appetite was associated with decrease in fat mass and BMI. HIV did not influence body composition regardless of gender.

The cross-sectional design employed in the present study limits the interpretation of finding in the present study to associations rather than causation. The findings suggest that energy and protein intakes, appetite, income, employment, and marital status are important predictors of body composition among women whereas tuberculosis and appetite are important among men. Energy intake in the presence of tuberculosis; reduced appetite among women and among men; tuberculosis among men; protein intake when

139

there is no income and when there is no employment were associated with negative prediction of body composition. Whereas protein intake when there is tuberculosis among women and being a single woman were associated with positive prediction of body composition. To our knowledge, this study is the first to show how energy and protein intakes influence body composition among women and among men and the interactions with tuberculosis, income, and employment. Of note, HIV infection did not provide any prediction of body composition regardless of gender. This corroborates with the claim in previous reports (Paton and Ng 2006; Mupere et al. 2010) that tuberculosis is the dominant factor in the wasting process even in the presence of HIV and that the disease in itself is associated with wasting (Rubin 1995). The major strengths in the present study is the composition of the study population with HIV positive and HIV negative adults with or without tuberculosis, allowing prediction of body composition in the presence or absence of tuberculosis and HIV infection.

This study found that the influence of energy and protein intakes on body composition differed by gender. A unit increase in energy intake was associated with a unit increase in

BMI among women; however, in the face of tuberculosis a unit increase in energy intake was associated with a unit decrease in BMI. Whereas protein intake among women with no income or unemployment were associated with a decrease in fat-free mass and fat mass, and protein intake in the presence of tuberculosis was associated with increase in

BMI. Energy and protein intakes among men did not influence body composition. A possible explanation to this gender difference in use of the dietary intake nutrient substrates rests on the differences in lipid and carbohydrate metabolism between men and

140

women (Tarnopolsky and Ruby 2001; Tarnopolsky 2000, 2008). Women use energy intake to build or maintain their fat content. Thus, an increase in BMI since BMI is a measure of fat content (Garrow and Webster 1985). However, during sub-maximal stress like lack of income or unemployment to purchase adequate nutrient substrates, they do oxidize away the lipid content sparing the lean tissue, and hence the decrease in BMI. It has reported in several reports that women oxidize more lipid and less of carbohydrate as metabolic substrates than men (Tarnopolsky 2000, 2008; Tarnopolsky and Ruby 2001).

However, in extreme stress such having tuberculosis disease, women may use protein substrates to maintain the BMI. For men, it appears they are in position to balance the use of nutrient substrates in building and maintaining both fat and fat-free mass content.

Thus, no effect is noticed in prediction models. Moreover, it has also been reported that at the time of tuberculosis diagnosis, the average weight group difference in men consisted of lean tissue and fat in equal proportions whereas in women, the average weight group difference consisted predominantly of fat mass (Mupere et al. 2010).

Findings in the present also show gender differences in the effect of reduced on body composition. Reduced appetite among women was associated with decrease in both fat and fat-free mass whereas among men, reduced appetite was associated with decrease in

BMI. This suggests that reduced appetite could be a predictor of both fat and fat-free mass among women whereas reduced appetite could be a predictor of BMI among men.

This probably reflects on the wasting process and the gender differences in body composition when anorexia sets in. Men waste away their high fat-free mass content to be in proportion with fat content whereas women, waste away their high fat content

141

sparing the low fat-free mass content. Although, BMI is monotonically related to

adiposity (Garrow and Webster 1985), it also correlates positively with the amount of fat-

free an individual has. Hence, the level of appetite may reflect on the body composition

for an individual. Women have little fat-free mass content, thus requiring daily energy intake to preserve physical function. Men in general display a higher absolute resting metabolic rate with associated energy expenditure than women because of their lager quantity of fat-free mass (Arciero, Goran, and Poehlman 1993). Thus, men require a good

appetite to maintain nutrient intake that can meet their metabolic demands.

Another possible explanation is the differences in sex hormones between men and

women that the sex hormones may play a role in regulation of symptoms associated

anorexia in tuberculosis. Sex hormones are physiologically associated with feeding

behavior (Geary 2001). Higher anorectic signals and earlier satiety have been reported in

men suffering from chronic illnesses, perhaps contributing to a different response pattern

to anorexigenic diseases such as cancer among men and women (Geary 2001). In animal

models, inflammation-induced anorexia has been reported to be more severe among male

rats (Lennie 2004) and previous reports suggest that estradiol and progesterone have

inhibitory effects on anorexia (Eckel 2004). Further, sex hormones or derivatives have

been used in treatment of eating disorders with positive results in improving appetite,

caloric intake, and nutritional and inflammatory status (Rammohan et al. 2005).

142

Findings in the present revealed that tuberculosis was a universal predictor of body composition. Having tuberculosis was associated a decrease in fat and fat-free mass and

BMI regardless of gender. Wasting is a recognized cardinal feature of tuberculosis. It is likely caused by a combination of reduction in appetite, leading to a decrease in energy intake, interacting with increased losses and altered metabolism as part of the inflammatory and immune responses (Paton et al. 2003; Macallan et al. 1998). Utilization of amino acids and protein synthesis may be inhibited due to the presence of pro- inflammatory cytokines (Macallan et al. 1998).

In conclusion, the present study has revealed that there are remarkable gender differences in how energy and protein intakes influence body composition, and there are important interactions with presence of tuberculosis, lack of income, and unemployment. HIV does not appear to influence nutrient intake on body composition. Further evaluation is needed to understand how these differences would influence body composition over time and survival.

Acknowledgements

We thank all study staff members of the Case Western Reserve University and Makerere

University research collaboration at the Tuberculosis Research Unit in the United States and in Uganda for their assistance; the faculty and staff at Case Western Reserve

University Department of Epidemiology and Biostatistics for the guidance in analyzing the project; and the Fogarty International Center, for the continued support.

143

This study was supported in part by the AIDS International Training Research Program,

Fogarty International Center, Grant No. TW000011, based at Case Western Reserve

University, Department of Epidemiology and Biostatistics. This work was part of Ezekiel

Mupere’s PhD thesis at Case Western Reserve University.

144

Table 6:1 Characteristics of the study population (n=131)

Characteristics All subjects (n=131)

Sex

Female n (%) 70 (53)

Male n (%) 61 (47)

Age in years

≤30 n (%) 94 (72)

>30 n (%) 37 (28)

HIV status

Negative n (%) 62 (47)

Positive n (%) 69 (53)

Tuberculosis

No n (%) 68 (52)

Yes n (%) 63 (48)

Education

None/primary level n (%) 69 (53)

Secondary level n (%) 62 (47)

145

Tribe

Muganda n (%) 57 (44)

Others n (%) 74 (56)

Marital status

Married n (%) 56 (43)

Single n (%) 37 (28)

Separated/divorced n (%) 38 (29)

Household number

One to two n (%) 40 (31)

>2 n (%) 91 (69)

Employed

No n (%) 54 (41)

Yes n (%) 77 (59)

Income

Not at all n (%) 47 (36)

Yes n (%) 84 (64)

Takes alcohol

146

No n (%) 98 (75)

Yes n (%) 33 (25)

Reduced appetite

No n (%) 89 (68)

Yes n (%) 42 (32)

147

Table 6:2 Nutrient intake characteristics of the study population (n=131)

Characteristics All subjects (n=131)

Energy intake in kcal [mean, SD)] 1820 (874)

Protein intake in g [mean, (SD)] 49.5 (32.1)

Body mass index in kg/m2 [mean, (SD)] 20.9 (3.8)

Fat-free mass in kg [mean, (SD)] 45.0 (7.0)

Fat-free mass index in kg/m2 [mean, (SD)] 16.8 (1.6)

Fat mass in kg [mean, (SD)] 10.4 (7.0)

Fat mass index in kg/m2 [mean, (SD)] 4.1 (3.1)

SD = standard deviation

148

Table 6:3 Spearman’s correlations between energy or protein intake and body mass index, fat or fat-free mass (n=131)

Women (n=70) Men (n=61) Characteristics Energy (kcal) Protein (g) Energy (kcal) Protein (g)

Body mass index (kg/m2) 0.28b 0.08 0.10 0.12

Fat-free mass (kg) 0.36b 0.13 0.20 0.18

Fat-free mass index (kg/m2) 0.28b 0.06 0.11 0.11

Fat mass (kg) 0.29b 0.10 0.04 0.11

Fat mass index (kg/m2) 0.25b 0.08 0.02 0.09 ap-value <0.001, bp-value <0.05.

149

Table 6:4 Predictors of fat-free mass in univariate models among HIV positive and

HIV negative adults with or without tuberculosis in Kampala, Uganda (n=131)

All (n=131)

Characteristics Unadjusted p-value

Estimate (SE)

Energy intake, kcal 0.001 (0.0002) <0.001

Protein intake, g 0.002 (0.001) 0.049

Older age >30 years -0.23 (0.30) 0.449

HIV positive -0.34 (0.27) 0.210

Tuberculosis -1.41 (0.24) <0.001

No/low education -0.41 (0.43) 0.339

Not Muganda by tribe -0.34 (0.27) 0.148

Single 0.84 (0.29) 0.005

Household >2 people -0.32 (0.30) 0.279

Unemployed -0.46 (0.28) 0.095

No income at all -0.57 (0.28) 0.045

Takes alcohol -0.03 (0.32) 0.928

150

Reduced appetite -1.55 (0.26) <0.001

151

Table 6:5 Predictors of fat-free mass in univariate models among HIV positive and

HIV negative adults with or without tuberculosis in Kampala, Uganda stratified by sex (n=131)

Women (n=70) Men (n=61)

Characteristics Unadjusted p- Unadjusted p-

Estimate (SE) value Estimate (SE) value

Energy intake, kcal 0.0004 (0.0002) 0.028 0.0002 (0.0003) 0.418

Protein intake, g 0.001 (0.001) 0.275 0.0002 (0.002) 0.910

Older age >30 years -0.02 (0.32) 0.937 -0.65 (0.45) 0.157

HIV positive -0.35 (0.28) 0.212 0.14 (0.42) 0.743

Tuberculosis -0.960 (0.25) 0.0003 -1.81 (0.35) <0.001

No/low education 0.12 (0.38) 0.754 -0.60 (0.84) 0.481

Not Muganda by tribe -0.41 (0.27) 0.144 -0.44 (0.42) 0.301

Single 0.16 (0.28) 0.573 0.98 (0.58) 0.096

Household >2 people 0.10 (0.32) 0.746 -0.28 (0.43) 0.525

Unemployed 0.11 (0.28) 0.692 -0.49 (0.46) 0.283

No income at all 0.18 (0.27) 0.508 -0.62 (0.52) 0.236

Takes alcohol 0.25 (0.31) 0.416 -0.21 (0.50) 0.682

152

Reduced appetite -1.15 (0.25) <0.001 -1.64 (0.44) 0.0004

153

Table 6:6 Predictors of fat-free mass in multivariable models among HIV positive and HIV negative adults with or without tuberculosis in urban Kampala, Uganda

(n=131)

Overall (n=131) Characteristic Estimate (SE) p-value

Energy intake, kcal 0.0004 (0.0002) 0.025

Protein intake, kcal 0.0004 (0.001) 0.756

HIV positive -0.40 (0.24) 0.096

Tuberculosis -0.86 (0.29) 0.004

Single 0.68 (0.24) 0.005

Household >2 people -0.39 (0.26) 0.138

No income 0.36 (0.43) 0.408

Reduced appetite -0.88 (0.31) 0.006

Protein*income -0.02 (0.01) 0.036

R Square 0.41 ap-value <0.001, bp-value <0.05. Variables for which p-value was < 0.50 in the unadjusted analyses were included in the multivariable model. Unemployment and other than Muganda tribe were dropped because they did not contribute much to the R-square.

154

Table 6:7 Predictors of fat-free mass in multivariable models among HIV positive and HIV negative adults with or without tuberculosis in Kampala, Uganda stratified according sex (n=131)

Stratified models

Characteristic Women (n=70) Men (n=61)

Estimate (SE) p-value Estimate (SE) p-value

Energy intake, kcal 0.0003 (0.0002) 0.094 0.0004 (0.0004) 0.313

Protein intake, kcal 0.002 (0.001) 0.289 -0.002 (0.003) 0.383

HIV positive -0.35 (0.25) 0.168 -0.26 (0.42) 0.546

Tuberculosis -0.59 (0.33) 0.078 -1.42 (0.45) 0.003

Single 0.58 (0.25) 0.022 0.47 (0.44) 0.295

Household >2 people -0.36 (0.29) 0.227 0.14 (0.45) 0.755

No income 0.80 (0.41) 0.056 -0.16 (0.89) 0.861

Reduced appetite -0.79 (0.33) 0.020 -0.77 (0.52) 0.144

Protein*income -0.02 (0.01) 0.027 -0.007 (0.01) 0.591

R Square 0.42 0.40 ap-value <0.001, bp-value <0.05. Variables for which p-value was < 0.50 in the unadjusted analyses were included in the multivariable model. Unemployment and other than Muganda tribe were dropped because they did not contribute much to the R-square.

155

Table 6:8 Predictors of fat mass in univariate models among HIV positive and HIV negative adults with or without tuberculosis in Kampala, Uganda (n=131)

All (n=131) Characteristics Unadjusted Estimate (SE) p-value

Energy intake, kcal -0.00002 (0.0003) 0.945

Protein intake, g -0.001 (0.002) 0.536

Older age >30 years 0.09 (0.060) 0.885

HIV positive 0.49 (0.54) 0.367

Tuberculosis -2.40 (0.50) <0.001

No/low education 0.11 (0.85) 0.90

Not Muganda by tribe -0.30 (0.55) 0.590

Single -1.10 (0.59) 0.064

Household >2 people 0.82 (0.59) 0.162

Unemployed 0.88 (0.55) 0.112

No income at all 1.46 (0.55) 0.009

Takes alcohol 0.75 (0.62) 0.23

Reduced appetite -2.02 (0.55) <0.001

156

Table 6:9 Predictors of fat mass in univariate models among HIV positive and HIV negative adults with or without tuberculosis in Kampala, Uganda stratified according to sex (n=131)

Characteristics Women (n=70) Men (n=61)

Unadjusted p-value Unadjusted p-value

Estimate (SE) Estimate (SE)

Energy intake, kcal 0.001 (0.0005) 0.020 -0.000004 (0.0002) 0.981

Protein intake, g 0.004 (0.003) 0.304 0.0003 (0.001) 0.821

Older age >30 years 0.53 (0.95) 0.581 0.07 (0.32) 0.829

HIV positive -0.43 (0.85) 0.619 0.45 (0.29) 0.132

Tuberculosis -3.57 (0.72) <0.001 -1.35 (0.24) <0.001

No/low education -0.76 (1.15) 0.508 -0.29 (0.60) 0.632

Not Muganda by tribe -0.29 (0.84) 0.728 -0.20 (0.30) 0.508

Single -0.11 (0.85) 0.900 -0.006 (0.42) 0.988

Household >2 people 0.62 (0.97) 0.524 0.06 (0.31) 0.849

Unemployed 0.48 (0.83) 0.563 -0.28 (0.33) 0.390

No income at all 0.75 (0.83) 0.374 -0.06 (0.38) 0.869

Takes alcohol 1.19 (0.93) 0.203 -0.23 (0.35) 0.519

157

Reduced appetite -3.47 (0.75) <0.001 -1.30 (0.30) <0.001

158

Table 6:10 Predictors of fat mass in multivariable models among HIV positive and

HIV negative adults with or without tuberculosis in Kampala, Uganda (n=131)

Overall model (n=131) Characteristic Estimate (SE) p-value

Energy intake, kcal -0.0005 (0.0008) 0.534

Protein intake, g -0.009 (0.005) 0.077

Tuberculosis -0.33 (1.31) 0.799

Household >2 people -2.06 (1.11) 0.066

Unemployed 1.66 (1.17) 0.158

No income -0.80 (1.50) 0.596

Reduced appetite -3.99 (1.29) 0.003

Energy*household >2 people 0.002 (0.001) 0.010

Energy*tuberculosis -0.003 (0.001) <0.001

Energy*reduced appetite 0.002 (0.001) 0.005

Energy*no income 0.002 (0.001) 0.021

Protein*tuberculosis -0.09 (0.02) <0.001

Protein*unemployement -0.05 (0.02) 0.003

159

R Square 0.41 ap-value <0.001, bp-value <0.05. Variables for which p-vale was < 0.50 in the unadjusted analyses were included in the multivariable model. Alcohol intake and being single were dropped because they did not contribute much to the R-square.

160

Table 6:11 Predictors of fat mass in multivariable models among HIV positive and

HIV negative adults with or without tuberculosis in Kampala, Uganda (n=131)

Stratified models

Women (n=70) Men (n=61) Characteristic p- Estimate (SE) p- Estimate (SE) value value

Energy intake, kcal -0.0002 (0.002) 0.880 -0.001 (0.0003) 0.150

Protein intake, g -0.01 (0.01) 0.539 0.001 (0.003) 0.710

Tuberculosis -0.37 (2.15) 0.866 -1.49 (0.89) 0.100

Household >2 people -2.86 (1.72) 0.102 0.33 (0.90) 0.715

Unemployed 2.67 (1.85) 0.155 -0.57 (0.78) 0.468

No income -1.95 (2.40) 0.420 -0.84 (1.00) 0.407

Reduced appetite -4.17 (1.94) 0.036 -1.90 (0.90) 0.040

Energy*household >2 people 0.002 (0.001) 0.080 0.0001 (0.0004) 0.885

Energy*tuberculosis -0.003 (0.001) 0.070 0.0002 (0.001) 0.751

Energy*reduced appetite 0.001 (0.001) 0.207 0.001 (0.0004) 0.134

Energy*no income 0.002 (0.001) 0.097 0.001 (0.005) 0.264

161

Protein*tuberculosis 0.07 (0.04) 0.100 -0.003 (0.02) 0.859

Protein*unemployement -0.08 (0.03) 0.010 -0.001 (0.01) 0.938

R Square 0.53 0.51 ap-value <0.001, bp-value <0.05. Variables for which p-vale was < 0.50 in the unadjusted analyses were included in the multivariable model. Alcohol intake and being single were dropped because they did not contribute much to the R-square.

162

Table 6:12 Predictors of body mass index in univariate models among HIV positive and HIV negative adults with or without tuberculosis in Uganda (n=131)

Characteristics All (n=131)

Unadjusted p-value

Estimate (SE)

Energy intake, kcal 0.001 (0.0004) 0.161

Protein intake, g 0.001 (0.003) 0.786

Older age >30 years -0.10 (0.74) 0.894

HIV positive 0.15 (0.66) 0.823

Tuberculosis -3.88 (0.57) <0.001

No/low education -0.27 (1.04) 0.797

Not Muganda by tribe -0.71 (0.67) 0.292

Single -0.22 (0.73) 0.762

Household >2 people 0.53 (0.72) 0.460

Unemployed 0.36 (0.67) 0.594

No income at all 0.89 (0.69) 0.198

Takes alcohol -0.76 (0.76)) 0.319

163

Reduced appetite -3.67 (0.63) <0.001

164

Table 6:13 Predictors of body mass index in univariate models among HIV positive and HIV negative adults with or without tuberculosis in Kampala, Uganda stratified according sex (n=131)

Women (n=70) Men (n=61)

Characteristics Unadjusted p-value Unadjusted p-value

Estimate (SE) Estimate (SE)

Energy intake, kcal 0.002 (0.001) 0.014 0.0002 (0.0004) 0.622

Protein intake, g 0.005 (0.004) 0.306 0.0005 (0.003) 0.864

Older age >30 years 0.59 (1.22) 0.633 -0.57 (0.72) 0.434

HIV positive -0.76 (1.09) 0.490 0.59 (0.67) 0.385

Tuberculosis -4.66 (0.91) <0.001 -3.15 (0.54) <0.001

No/low education -0.57 (1.47) 0.698 -0.89 (1.35) 0.514

Not Muganda by tribe -0.72 (1.08) 0.505 -0.64 (0.67) 0.347

Single 0.08 (1.09) 0.945 0.97 (0.94) 0.306

Household >2 people 0.82 (1.24) 0.514 -0.22 (0.69) 0.755

Unemployed 0.52 (1.07) 0.625 -78 (0.73) 0.293

No income at all 0.96 (1.06) 0.372 -0.69 (0.84) 0.418

Takes alcohol 1.53 (1.19) 0.202 -0.44 (0.80) 0.587

165

Reduced appetite -4.78 (0.934) <0.001 -2.95 (0.68) <0.001

166

Table 6:14 Predictors of body mass index in multivariable models among HIV positive and HIV negative adults with or without tuberculosis in urban Uganda

(n=131)

Characteristic Overall model (n=131)

Estimate (SE) p-value

Energy intake, kcal 0.001 (0.001) 0.193

Protein intake, g -0.01 (0.01) 0.105

Tuberculosis -1.07 (1.52) 0.482

No income 1.01 (0.57) 0.079

Reduced appetite -4.93 (1.54) 0.002

Energy*tuberculosis -0.002 (0.001) 0.013

Energy*appetite 0.002 (0.001) 0.026

Protein*tuberculosis 0.06 (0.03) 0.033

R Square 0.39 ap-value <0.001, bp-value <0.05. Variables for which p-vale was < 0.50 in the unadjusted analyses were included in the multivariable model. Being other than Muganda tribe was dropped because it did not contribute much to the R-square.

167

Table 6:15 Predictors of body mass index in multivariable models among HIV positive and HIV negative adults with or without tuberculosis in Kampala, Uganda

(n=131)

Stratified models

Characteristic Women (n=70) Men (n=61)

Estimate (SE) p-value Estimate (SE) p-value

Energy intake, kcal 0.003 (0.001) 0.028 -0.001 (0.001) 0.311

Protein intake, g -0.02 (0.01) 0.119 0.003 (0.01) 0.622

Tuberculosis -0.48 (2.27) 0.834 -3.34 (1.76) 0.063

No income 1.01 (0.87) 0.250 -0.69 (0.67) 0.306

Reduced appetite -4.21 (2.25) 0.066 -4.3 (1.94) 0.030

Energy*tuberculosis -0.004 (0.001) 0.008 0.001 (0.001) 0.238

Energy*appetite 0.001 (0.001) 0.580 0.001 (0.001) 0.105

Protein*tuberculosis 0.10 (0.05) 0.039 -0.03 (0.03) 0.236

R Square 0.48 0.47 ap-value <0.001, bp-value <0.05. Variables for which p-vale was < 0.50 in the unadjusted analyses were included in the multivariable model. Being other than Muganda tribe was dropped because it did not contribute much to the R-square.

168

CHAPTER 7

BODY WASTING AND DIETARY INTAKE AMONG HIV POSITIVE AND HIV

NEGATIVE ADULTS WITH OR WITHOUT TUBERCULOSIS IN URBAN

UGANDA, KAMPALA

169

Abstract

Background The effects of tuberculosis and effect of HIV on dietary intake have not been well described. We aimed to establish the 1) independent effects of tuberculosis and

HIV infection on dietary intake, 2) relationship between dietary intake and body wasting

as measured by height-normalized fat-free mass (FFMI) and body mass (BMI) indices,

and 3) relationship between dietary intake and tuberculosis disease severity among HIV

positive and HIV negative adults with/or without tuberculosis in urban Uganda.

Methods In a cross-sectional study of 131 adults who were screened for active

tuberculosis and HIV infection, FFMI, BMI, and 24-hour dietary intake recall were

assessed.

Results Tuberculosis patients that had moderate/or severe clinical disease had lower

dietary intakes for energy, protein, total fat, carbohydrate, calcium, vitamin A, and folate

compared to patients with mild disease. Both men and women had comparable dietary

intake among patients with tuberculosis regardless of HIV status whereas HIV negative

women had reduced energy, protein, and folate intake among individuals without

tuberculosis compared to men. Tuberculosis patients with fat-free mass wasting or those

with reduced BMI had comparable nutrient intakes with counterparts that had normal fat- free mass or normal BMI.

Conclusion The study revealed that dietary intake at the time of diagnosis was influenced by tuberculosis disease severity, but not tuberculosis disease or HIV status and in the absence of tuberculosis was influenced by gender. Nutritional counseling and

170

supplementation, early treatment and prevention of tuberculosis are needed to improve dietary intake in populations of sub-Saharan Africa.

171

Background

Body wasting and malnutrition is endemic in sub-Saharan Africa (Muller and Krawinkel

2005), which also bears the highest burden of tuberculosis patients with human

immunodeficiency virus (HIV) co-infection (Lawn and Churchyard 2009). An estimated

1.37 million new cases of tuberculosis with HIV co-infection occurred in 2007; 79% of which were from sub-Saharan Africa. Co-infection with tuberculosis and HIV poses an extra burden to the pathophysiology of body wasting, exacerbating the wasting process seen in tuberculosis or HIV infection alone (Lucas et al. 1994; Macallan 1999).

Moreover, co-infection and malnutrition have deleterious interactions. Co-infected patients with malnutrition have high risk of morbidity and mortality (Zachariah et al.

2002; Lucas et al. 1994; Duarte et al. 2009) and tuberculosis is the leading cause of death in co-infected patients in tuberculosis endemic countries, including those with free access to antiretroviral therapy (Saraceni et al. 2008).

Co-infection may lead to poor appetite with decreased nutrient intake, which may interact with altered metabolism associated with both infections as part of the immune and inflammatory responses (Paton et al. 2003) leading to exacerbation of the existing body wasting. Yet poor nutritional status is associated with risk of tuberculosis relapse (Khan et al. 2006) in addition to morbidity and mortality. The goal of nutritional assessment and nutritional support is to intervene early and to preserve lean tissue or fat-free mass body compartment from further wasting because disproportionate loss of fat-free mass is associated with morbidity and mortality (Heitmann et al. 2000). Fat-free mass is a marker

172

of body wasting and malnutrition, because it is a consequence of negative imbalance

between energy (and protein) needs and dietary intake that occurs for more than a few

days.

Assessment of dietary intake is essential in nutritional management and in understanding

of body wasting and malnutrition. Despite the high burden of malnutrition and the high

burden of co-infection with associated body wasting, assessment of dietary intake is often

neglected in clinical practice and national tuberculosis programs. Thus, the effects of

tuberculosis and effect of HIV among co-infected patients on dietary intake have not been well described. The present cross-sectional study was conducted to establish the 1) independent effects of tuberculosis and HIV infection on dietary intake, 2) relationship between dietary intake and body wasting, and 3) relationship between dietary intake and

TB disease severity. The study was conducted among HIV positive and HIV negative adults with/or without active tuberculosis in urban Kampala, Uganda.

Subjects and methods

In a cross-sectional study, 132 participants 18 years or older residing in Kampala district or 20 km from the study site if residence was outside Kampala in Uganda were enrolled.

One participant was excluded from the analysis because of prior TB treatment. The study was conducted at the National tuberculosis and Leprosy Program (NTLP) Clinic of the national tertiary teaching hospital, Mulago complex between November 2007 and March

2008. Of the 131 participants who were included in the analysis, 63 were tuberculosis

173

patients who were recruited at the Mulago NTLP Clinic; 38 were HIV positive without

tuberculosis and recruited at the Infectious Disease Institute Clinic (IDI) located 500

meters from the Mulago NTLP Clinic; and 30 were HIV negative without tuberculosis

from the community where enrolled tuberculosis patients resided. The institutional

review boards at Case Western Reserve University and Joint Clinical Research Center

approved the study, with final approval by the Uganda National Council for Science and

Technology. All participants provided written informed consent to the study.

All subjects in the study were given appropriate pre- and post-test HIV counseling and

AIDS education. HIV-1 infection was diagnosed on the basis of a positive enzyme-linked immunosorbent assay for HIV-1 antibodies (Recombigen; Cambridge Biotech,

Cambridge, MA). At enrollment, basic demographic information and a medical history were collected, and a standardized physical examination was conducted by a medical officer. Active pulmonary tuberculosis was confirmed by sputum smear microscopy and culture. Patients with active tuberculosis were treated with standard four-drug chemotherapy for tuberculosis per guidelines of the Ugandan Ministry of Health.

Similarly, HIV positive patients eligible for antiretroviral therapy were started on treatment and cotrimoxazole prophylaxis at the IDI clinic.

We defined body wasting of participants using body mass index (BMI) and height- normalized indices (adjusted for height2) of body composition that partition BMI into fat-

free mass index (FFMI) and fat mass index (FMI) (Kyle, Piccoli, and Pichard 2003;

174

VanItallie et al. 1990; Schutz, Kyle, and Pichard 2002) to establish the body wasting status of participants. The FFMI and FMI have the advantages of compensating for differences in height and age (Kyle, Genton, and Pichard 2002). Also, the use of the

FFMI and FMI eliminates some of the differences between population groups. We defined body wasting as patients having the low fat-free mass index (FFMI) and the low body fat mass index (FMI) as previously recommended (Kyle, Piccoli, and Pichard 2003;

VanItallie et al. 1990). The low FFMI of 16.7 (kg/m2) for men and 14.6 (kg/m2) for women and the low FMI of 1.8 (kg/m2) for men and 3.9 (kg/m2) for women corresponds to a BMI of <18.5 kg/m2, the WHO cutoff for malnutrition (World Health Organ Tech

Rep 1995) among adults.

100g. Height was measured to the nearest cm by calibrated standing height using a stadiometer. All anthropometric measurement values were the mean of duplicates. The single-frequency bioelectrical impedance analyzer (BIA Detroit, MI, RJL Systems) performing at 50 kHz and 800 mA was used for BIA measures with detecting electrodes placed on the wrist and ankle and signal introduction electrodes placed on the first joint of the middle finger and behind the middle toe.

175

Before performing measurements on each subject, the BIA instrument was calibrated using the manufacturer’s recalibration device. The resistance and reactance were based on measures of a series circuit (Kotler et al. 1996). BIA measurements were performed in duplicate for each subject. The analyzer was calibrated monthly. Fat-free mass was calculated from BIA measurements using equations that were previously cross-validated in a sample of patients (white, black and Hispanic) with and without HIV infection

(Kotler et al. 1996) and have been applied elsewhere in African studies (Shah et al. 2001;

Van Lettow et al. 2004; Villamor et al. 2006). Fat mass was calculated as body weight minus fat-free mass.

We classified tuberculosis patients into two categories of clinical disease severity as follows: mild and moderate/or severe tuberculosis using the TBscore. HIV negative participants without tuberculosis were classified as no disease category. The clinical tuberculosis score is a low-cost tool that has been developed recently and validated to assess severity of clinical TB disease in resource limited settings (Wejse et al. 2008). The score has a maximum of 13 clinical variables each scoring 1 point as follows: self- reported symptoms of cough, hemoptysis, dyspnea, chest pain, and night sweating; anemic conjunctivae, tachycardia, positive finding at lung auscultation, axillary temperature >37.00C, body mass index (BMI) <18, BMI <16, middle upper arm circumference (MUAC) <220 mm, MUAC <200. Clinical severity categories were generated using the following cutoffs: mild 0 – 5 and moderate/severe > 5. Higher scores are associated with severe disease.

176

The dietary intake assessment was made using a single 24-hour dietary recall with open

ended questions. The questionnaire was pre-tested by administering it to 8 individuals selected randomly from the neighboring community to the study site. The assessment was conducted by four trained study nursing staffs and supervised a nutritionist using local food photographs, portion-size images, and volumetric vessels to increase the accuracy of the recall from the previous 24 hours. The nutritive value of raw ingredients was computed using the East African food composition table database whose database was imported into the NutriSurvey software (http://www.nutrisurvey.de) to easy the computations. When the East African food composition table was found deficient in certain food items, the United States Department of Agriculture database and the African composition table were used.

Analysis

All study participants in the analysis were categorized into 4 mutually independents groups: HIV positive patients with and without tuberculosis disease, HIV negative

patients with and without tuberculosis. Measures of central tendency and variability were

compared between women and men across 4 mutually exclusive groups, between patients

with and patients without reduced FFMI, between patients with and without reduced

BMI, and between patients with mild and with moderate/severe disease using Wilcoxon-

Mann Whitney test for average weight, height, BMI, fat and fat-free mass, FFMI, FMI,

and nutrient intake parameters due to lack of normality.

177

We made comparisons using analysis of variance between HIV positive patients with

tuberculosis and HIV positive patients without tuberculosis, between HIV negative

patients with tuberculosis and HIV negative individuals without tuberculosis, between

HIV positive patients with tuberculosis and HIV negative patients with tuberculosis, and

between HIV positive individuals without tuberculosis and HIV negative individuals

without tuberculosis to understand the independent effects of tuberculosis and HIV

infection on dietary intake and nutritional status. We used Bonferroni to adjust for

multiple comparisons. We compared proportions of nutrient inadequacy between women

and men across the 4 mutually exclusive groups, between patients with and without

reduced FFMI, between patients with and without reduced BMI, and between patients

with mild and patients with moderate/severe clinical tuberculosis using chi-square and

Fischer’s exact tests. Fischer’s exact test was used where expected counts were less than

5. A p-value of <0.05 was considered significant in all analyses except for multiple comparisons which was set at <0.01. All analyses were performed using SAS version 9.2

(Cary software, North Carolina SAS Institute Inc 2004).

Results

178

tuberculosis; men were significantly older than women 34.7 ± 6.5 (SD) versus 29.7 ± 8.4

(SD), respectively.

Among HIV positive and HIV negative patients with tuberculosis, men and women had

comparable BMI whereas among HIV positive and HIV negative individuals without

tuberculosis, women had significantly higher BMI compared to men (Table 7:1 and 7:2).

Men had significantly higher fat-free mass and significantly lower fat mass compared to women regardless of tuberculosis and HIV status. Similarly, men had a higher magnitude of height-normalized FFMI and a lower height-normalized FMI compared to women regardless of tuberculosis and HIV status. Height-normalized FFMI and FMI eliminate differences in fat and fat-free mass associated with height (Baumgartner et al. 1998). Men with tuberculosis had significantly higher proportion of individuals with reduced FFMI compared to women regardless of HIV status (Table 7:1 and 7:2). For example among

HIV positive patients with tuberculosis, 7 out of 10 (70%) for men had reduced FFMI compared to 2 out of 21 (10%) for women (p<0.001). The clinical TB severity score was comparable between men and women regardless of HIV status. Regardless of gender on multiple comparisons, we found no differences in average BMI, fat and fat-free mass, and tuberculosis clinical severity score between HIV positive patients with tuberculosis and

HIV positive patients without tuberculosis, between HIV negative patients with tuberculosis and HIV negative individuals without TB, between HIV positive patients with tuberculosis and HIV negative patients with tuberculosis, and between HIV positive individuals without tuberculosis and HIV negative individuals without tuberculosis.

179

In general, there were no significant differences in average 24-hour dietary intake recall

between men and women for most nutrients except average energy, protein, total fat, and

folate among HIV negative individuals without tuberculosis; dietary fiber among HIV

positive individuals without tuberculosis; magnesium and zinc intake among HIV

negative patients with tuberculosis (Tables 7:3 and 7:4). Of note, however, men had a

high magnitude of average dietary intake for most nutrients, for example energy intake

was 2380 ± 703 for men versus 1598 ± 567 for women among HIV negative individuals

without tuberculosis, 2027 ± 1111 for men versus 1437 ± 920 for women among HIV

positive patients with tuberculosis (Table 7:3 and 7:4).

On average, BMI, fat mass, and height-normalized FMI were significantly lower among

tuberculosis patients with reduced fat-free mass, reduced BMI, and moderate/severe

TBscore compared to patients with normal fat-free mass, normal BMI, and mild TBscore.

However, the average severity TBscore was significantly higher among patients with reduced fat-free mass, reduced BMI, and moderate/severe TBscore (Table 7:5 and 7:6).

The average fat-free mass and height-normalized FFMI were similar between TB patients

with mild and patients with moderate/severe TBscore. However, patients with mild

TBscore had significantly higher proportion of individuals with reduced fat-free mass (16

out of 39 (41%)) compared to patients with moderate/severe TBscore (4 out of 24 (17%),

p <0.05) (Table 7:5 and 7:6).

180

Tuberculosis patients with moderate/severe TBscore had significantly lower average 24- hour dietary intake recall for average energy, protein, total fat, dietary fiber, calcium, vitamin A, and folate compared to patients with mild TBscore (Table 7:7 and 7:8). For example, energy intake among patients with moderate/severe TBscore was 1460 ± 722 compared to 2215 ± 940 among patients with mild TBscore. However, there were no significant differences in average 24-hour dietary intake recall between tuberculosis patients with reduced fat-free mass or reduced BMI and patients with normal fat-free mass or normal BMI (Table 7:7 and 7:8).

Discussion

In this cross-sectional study, we aimed to establish the relationship between dietary intake and body wasting; the independent effects of tuberculosis and HIV infection, and tuberculosis disease severity on dietary intake. In a study population of 131 HIV positive and HIV negative adults with/or without active tuberculosis from urban Uganda, we found that the 24-hour dietary intake recall varied by severity of tuberculosis disease, but not tuberculosis disease or HIV status. In the absence of tuberculosis, dietary intake varied by gender. The dietary intake differed by severity of clinical tuberculosis disease categories of mild and moderate/or severe disease. Men and women with tuberculosis had similar dietary intake regardless of HIV status. HIV negative women without tuberculosis had lower levels of dietary intake for energy, protein, and folate compared to HIV negative men without tuberculosis. Dietary intake did not differ by severity of nutritional status.

181

Findings of the present study suggest that in the face of tuberculosis disease, dietary

intake is affected by severity of disease but not HIV infection; and there is no association

with body wasting. While in the absence of tuberculosis, dietary intake is affected by

gender, and not HIV infection. Tuberculosis patients that had moderate/or severe clinical

disease had lower dietary intakes for energy, protein, total fat, carbohydrate, calcium,

vitamin A, and folate compared to patients with mild disease. Both men and women had comparable dietary intake among patients with tuberculosis regardless of HIV status whereas HIV negative women had reduced energy, protein, and folate intake among individuals without tuberculosis compared to men. Tuberculosis patients with fat-free

mass wasting or those with reduced BMI had comparable nutrient intakes with

counterparts that had normal fat-free mass or normal BMI. To our knowledge, this is the

first study to have evaluated the effect of tuberculosis disease severity on dietary intake

and whether there are differences in dietary intake between HIV positive and HIV

negative adults with tuberculosis. Our results are consistent with recent findings from

India in which the study (Swaminathan et al. 2008) found comparable nutrient dietary

intakes between HIV positive patients with tuberculosis and HIV positive individuals

without tuberculosis. This study however, had no comparable group of HIV negative

patients with tuberculosis to establish the synergist effect of tuberculosis and HIV

infection. The strengths of our study hinges on the full panel of HIV positive and HIV

negative adults with/or without tuberculosis. However, the study limited findings are

limited by the cross-sectional nature of the design that the associations are not causal

between differences between groups.

182

The present study demonstrated that dietary intake at the time of tuberculosis diagnosis was influenced by disease severity and in populations without tuberculosis was influenced by gender. Dietary intake differed by tuberculosis disease severity and not by tuberculosis disease or HIV status. While in populations without tuberculosis, nutrient intakes differed by gender particularly among HIV negative individuals. Women had reduced intakes for energy, protein, and folate. Despite high proportions of patients with body wasting as assessed by reduced fat-free mass (32% (20/63) with reduced fat-free mass) or BMI (56% (35/63) with reduced BMI), nutrient intakes were not associated with body wasting. Nutrient intakes were similar between individuals with wasting and those without. Two potential reasons can explain the findings in the present study. First, the disparity in reduced appetite between patients with mild clinical disease and those with moderate/or severe disease. Patients with moderate/or severe clinical disease present with marked low appetite that impends the nutrient intake other than barriers to food access.

Further, the comparability of appetite level among men and women with tuberculosis, and among patients with/or without body wasting explains the gender and the wasting status similarity in nutrient intake. One can postulate that tuberculosis disease affects the appetite level equally in men and in women regardless of body wasting status, thus, compromising nutrient intake at nearly the same rate regardless of food access barriers.

Thus, the gender differences in body composition during tuberculosis as revealed in the present and in previous reports (Mupere et al. 2010) could be explained by the differences in altered metabolism. Men generally, experience higher metabolic rate because of the larger quantities of fat-free mass compared to women (Arciero, Goran, and

183

Poehlman 1993) and during tuberculosis, this process is probably pronounced with associated wasting.

The gender differences in dietary intake among HIV negative individuals without tuberculosis could be explained by the cultural factors that may compromise intake among women. For example, unequal distribution of food within households (Carloni

1981; de Hartog A.P 1972), or men may have the opportunity to eat a wider variety or better quality foods outside the home, such as cafes or local restaurants (Holmboe-

Ottesen G and Wandel M 1991). The unequal distribution of food within households result from several factors such as women may be trained to show restraint in eating, to give the best foods to men, or to allow others in the family to eat first (Lado 1992;

O'Laughlin B 1974; Rosenberg E.M 1980; Dey J 1981).

To conclude, the study revealed that dietary intake at the time of diagnosis was influenced by tuberculosis disease severity, but not tuberculosis disease or HIV status and in the absence of tuberculosis was influenced by gender. Nutritional counseling and supplementation, early treatment and prevention of tuberculosis are needed to improve dietary intake in population of sub-Saharan Africa.

184

Acknowledgements

We thank all study staff members of the Case Western Reserve University and Makerere

University research collaboration at the Tuberculosis Research Unit in the United States and in Uganda for their assistance; the faculty of staff at Case Western Reserve

University Department of Epidemiology and Biostatistics for the guidance in analyzing the project; and the Fogarty International Center, for the continued support.

This study was supported in part by the AIDS International Training Research Program,

Fogarty International Center, Grant No. TW000011, based at Case Western Reserve

University, Department of Epidemiology and Biostatistics. This work was part of Ezekiel

Mupere’s PhD thesis at Case Western Reserve University.

185

Table 7:1 Select characteristics among HIV positive and HIV negative adults with tuberculosis

Characteristic HIV positive with TB HIV negative with TB

[mean, (SD)] (n=31) (n=32)

Men Women Men Women

(n=10) (n=21) (n=18) (14)

Age in yr 30.9 (4.6) 29.2 (5.9) 26.0 (7.3) 26.3 (4.6)

BMI kg/m2 18.4 (1.7) 18.6 (3.0) 18.2 (2.0) 20.3 (4.3)

BMI

<18.5 kg/m2 (%) 7 (70) 12 (57) 11 (61) 5 (36)

≥18.5 kg/m2 (%) 3 (30) 9 (43) 7 (39) 9 (64)

Fat-free mass in kg 49.1 (4.5) 38.8 (3.2)a 49.1 (4.7) 39.9 (5.6)a

FFMI in kg/m2 16.6 (1.3) 15.4 (0.9)b 16.6 (1.5) 16.1 (0.9)

FFMI

<16.7 for M, <14.6 for W (%) 7 (70) 2 (10)a 11 (61) 0 (0)a

≥16.7 for M, ≥14.6 for W (%) 3 (30) 19 (90) 7 (39) 14 (100)

Fat mass in kg 5.2 (1.8) 8.3 (5.2) 4.8 (2.3) 10.5 (6.6)b

186

FMI in kg/m2 1.8 (0.6) 3.3 (2.1) 1.6 (0.8) 4.6 (3.6)a

FMI

<1.8 for M, <3.9 for W (%) 5 (50) 14 (67) 11 (61) 8 (57)

≥1.8 for M, ≥3.9 for W (%) 5 (50) 7 (33) 7 (39) 6 (43)

Severity TBscore 6.5 (1.5) 5.6 (2.6) 6.8 (2.3) 5.9 (2.4)

Severity TBscore category

Mild ≤5 (%) 2 (20) 10 (48) 6 (33) 6 (43)

Moderate/severe >5 (%) 8 (80) 11 (52) 12 (67) 8 (57) ap-value <0.001, bp-value <0.05. BMI = body mass index, FFMI = fat-free mass index, FMI = fat mass index, SD = standard deviation, W = women, M = men.

187

Table 7:2 Select characteristics among HIV positive and HIV negative adults without tuberculosis

Characteristic HIV positive, no TB HIV negative, no TB

[mean, (SD)] (n=38) (n=30)

Men Women Men Women

(n=17) (21) (n=16) (14)

Age in yr 34.7 (6.5) 29.7 (8.4)b 22.4 (3.2) 24.3 (5.4)

BMI kg/m2 21.2 (2.2) 24.2 (4.6)b 21.6 (2.3) 23.7 (2.8)b

BMI

<18.5 kg/m2 (%) 2 (12) 5 (24) 1 (6) 1 (7)

≥18.5 kg/m2 (%) 15 (88) 16 (76) 15 (94) 13 (93)

Fat-free mass in kg 53.0 (4.8) 39.8 (3.0)a 51.9 (3.9) 42.5 (4.9)a

FFMI in kg/m2 18.3 (1.2) 16.6 (1.2)a 18.6 (1.4) 16.6 (1.1)a

FFMI

<16.7 for M, <14.6 for W (%) 2 (12) 0 (0) 1 (6) 1 (7)

≥16.7 for M, ≥14.6 for W (%) 15 (88) 21 (100) 15 (94) 13 (93)

Fat mass in kg 9.1 (3.2) 17.8 (8.2)b 7.9 (2.5) 18.2 (5.4)a

188

FMI in kg/m2 3.2 (1.2) 7.6 (3.7)a 2.9 (1.0) 7.1 (2.1)a

FMI

<1.8 for M, <3.9 for W (%) 3 (18) 5 (24) 1 (6) 1 (7)

≥1.8 for M, ≥3.9 for W (%) 14 (82) 16 (76) 15 (94) 13 (93) ap-value <0.001, bp-value <0.05. BMI = body mass index, FFMI = fat-free mass index, FMI = fat mass index, SD = standard deviation, W = women, M = men.

189

Table 7:3 Dietary intake of 24-hour recall among HIV positive and HIV negative adults with tuberculosis

Characteristic HIV positive with TB HIV negative with TB

[mean, (SD)]1 (n=31) (n=32)

Men (n=10) Women (n=21) Men (n=18) Women (n=14)

Energy (kcal) 2027 (1111) 1437 (920) 2094 (831) 1567 (511)

Protein (g) 60.6 (48.6) 49.5 (44.2) 56.0 (29.5) 42.8 (22.3)

Total fat (g) 59.5 (60.2) 32.4 (38.4) 56.5 (38.9) 39.8 (24.9)

Carbohydrate (g) 303 (158) 245 (141) 348 (139) 269 (82)

Dietary fiber (g) 42.7 (32.9) 30.2 (22.5) 41.4 (23.2) 39.8 (25.0)

Protein, % energy 11.9 (6.5) 13.2 (8.5) 10.7 (2.1) 10.7 (3.7)

Fat, % energy 23.4 (13.8) 15.9 (9.4) 22.5 (10.4) 21.1 (8.5)

CHO, % energy 64.7 (16.5) 70.7 (13.8) 66.9 (10.2) 68.3 (9.9)

Calcium (mg) 978 (1532) 558 (1029) 297 (319) 264 (200)

Magnesium (mg) 267 (267) 261 (250) 302 (160) 176 (134)b

Zinc (mg) 7.7 (5.9) 5.6 (4.5) 6.9 (3.2) 4.4 (2.1)b

Iron (mg) 11.9 (12.5) 11.5 (12.6) 11.8 (8.2) 9.0 (5.4)

190

Vitamin A (RE) 469 (568) 493 (108) 545 (835) 779 (745)

Ascorbic acid

(mg) 105 (64) 108 (140) 142 (159) 130 (80)

Vitamin D (μg) 1.1 (2.3) 1.0 (1.5) 0.6 (1.1) 0.6 (0.9)

Folate (μg) 300 (221) 315 (278) 346 (257) 375 (183) ap-value <0.001, bp-value <0.05. CHO = carbohydrate, 1Characteristic values are means ± standard deviation (SD). TB = tuberculosis

191

Table 7:4 Dietary intake of 24-hour recall among HIV positive and HIV negative adults without tuberculosis

Characteristic HIV positive, no TB HIV negative, no TB

[mean, (SD)]1 (n=38) (n=30)

Men (n=17) Women (21) Men (n=16) Women (14)

Energy (kcal) 1954 (699) 1651 (1105) 2380 (703) 1598 (567)a

Protein (g) 56.9 (34.5) 38.3 (24.8) 60.4 (21) 35.9 (16.3)a

Total fat (g) 47.2 (28.5) 34.0 (20.0) 60.7 (25.1) 38.2 (23.3)b

Carbohydrate (g) 333 (125) 301 (242) 416 (119) 274 (124)

Dietary fiber (g) 45.4 (27.8) 26.3 (16.8)b 40.1 (18.6) 23.3 (14.6)

Protein, % energy 11.2 (3.3) 9.7 (3.3) 9.8 (1.9) 8.9 (2.5)

Fat, % energy 20.8 (8.9) 19.9 (9.4) 21.4 (4.4) 20.5 (9.2)

CHO, % energy 68.0 (10.4) 70.6 (11.0) 68.7 (5.3) 67.7 (14.3)

Calcium (mg) 415 (414) 263 (404) 417 (454) 346 (406)

Magnesium (mg) 238 (204) 251 (336) 278 (181) 190 (150)

Zinc (mg) 7.1 (4.3) 5.7 (4.8) 7.2 (3.3) 3.8 (1.8)

Iron (mg) 11.1 (7.7) 11.5 (14.2) 12.4 (7.0) 7.0 (5.3)

192

Vitamin A (RE) 188 (238) 959 (1794) 936 (2716) 285 (314)

Ascorbic acid (mg) 129 (103) 142 (142) 144 (87) 86 (51)

Vitamin D (μg) 0.6 (1.2) 0.1 (0.4) 0.5 (0.8) 0.6 (0.8)

Folate (μg) 362 (174) 312 (190) 513 (218) 293 (148)b ap-value <0.001, CHO = carbohydrate, bp-value <0.05. 1Characteristic values are means ± standard deviation (SD). TB = tuberculosis

193

Table 7:5 Select characteristics among patients with/without wasting (n=63)

Fat-free mass BMI Characteristic Not wasted Wasted Not wasted Wasted [mean, (SD)] (n=43) (n=20) (n=28) (n=35)

Age in yr 28.1 (5.4) 27.4 (8.0) 28.0 (5.7) 27.8 (6.7)

BMI kg/m2 19.8 (3.2) 16.9 (1.2)a 21.3 (2.8) 16.9 (1.3)a

BMI

<18.5 kg/m2 (%) 16 (37) 19 (95)a - -

≥18.5 kg/m2 (%) 27 (63) 1 (5) - -

Fat-free mass in kg 42.2 (6.5) 46.2 (5.5)b 43.8 (7.3) 43.1 (5.7)

FFMI in kg/m2 16.3 (1.4) 15.6 (0.8)b 16.8 (1.4) 15.5 (0.8)a

FFMI

<16.7 for M, <14.6 for W (%) - - 1 (4) 19 (54)a

≥16.7 for M, ≥14.6 for W (%) - - 27 (96) 16 (46)

Fat mass in kg 8.8 (5.3) 4.1 (2.0)a 10.9 (5.3) 4.4 (2.2)a

FMI in kg/m2 3.6 (2.6) 1.4 (0.6)a 4.4 (2.8) 1.6 (0.9)a

FMI

194

<1.8 for M, <3.9 for W (%) 20 (47) 5 (25) 7 (25) 31 (89)a

≥1.8 for M, ≥3.9 for W (%) 23 (53) 15 (75) 21 (75) 4 (11)

Severity TBscore 5.6 (2.2) 8.0 (7.5)b 5.0 (2.1) 7.1 (2.1)a

ap-value <0.001, bp-value <0.05. FFMI = fat-free mass index, FMI = fat mass index, BMI = body mass index, W = women, M = men, and SD = standard deviation. Fat-free mass wasting = FFMI <16.7 kg/m2 for men and <14.6 kg/m2 for women, BMI wasting = <18.5 kg/m2 for men and women.

195

Table 7:6 Select characteristics among patients with/without severity of clinical tuberculosis

Severity TBscore Characteristic Mild ≤5 Moderate/severe >5 [mean, (SD)] (n=39) (n=24)

Age in yr 28.0 (6.8) 27.7 (6.0)

BMI kg/m2 20.2 (3.8) 18.0 (2.1)b

BMI

<18.5 kg/m2 (%) 24 (62) 11 (46)

≥18.5 kg/m2 (%) 15 (38) 13 (54)

Fat-free mass in kg 43.2 (7.0) 43.6 (6.1)

FFMI in kg/m2 16.3 (1.2) 15.9 (1.3)

FFMI

<16.7 for M, <14.6 for W (%) 16 (41) 4 (17)b

≥16.7 for M, ≥14.6 for W (%) 23 (59) 20 (83)

Fat mass in kg 9.6 (6.2) 5.9 (3.5)b

FMI in kg/m2 4.0 (3.2) 2.2 (1.4)b

196

FMI

<1.8 for M, <3.9 for W (%) 26 (67) 12 (50)

≥1.8 for M, ≥3.9 for W (%) 13 (33) 12 (50)

Severity TBscore 3.9 (1.3) 7.6 (1.6)a

197

Table 7:7 Dietary intake of 24-hour recall among tuberculosis patients with/without body wasting (n=63)

Fat-free mass BMI Characteristic Not wasted Wasted Not wasted Wasted [mean, (SD)]1 (n=43) (n=20) (n=28) (n=35)

Energy (kcal) 1680 (790) 1893 (1070) 1727 (702) 1763 (1018)

Protein (g) 50.3 (33) 54.5 (44.3) 52.4 (36.8) 51.0 (37.3)

Total fat (g) 39.4 (32.3) 57.8 (54.2) 39.1 (22.8) 50.2 (50.9)

Carbohydrate (g) 288 (138) 290 (137) 296 (140) 283 (135)

Dietary fiber (g) 37.6 (24.4) 37.4 (27.0) 37.7 (25.1) 37.3 (25.4)

Protein, % energy 12.0 (6.5) 11.1 (4.5) 12.1 (6.6) 11.5 (5.4)

Fat, % energy 18.8 (9.6) 23.1 (12.0) 19.8 (9.0) 20.5 (11.7)

CHO, % energy 69.2 (12.2) 65.9 (12.8) 68.3 (11.6) 68.1 (13.3)

Calcium (mg) 448 (759) 565 (1125) 529 (895) 450 (885)

Magnesium (mg) 244 (193) 279 (242) 254 (161) 255 (242)

Zinc (mg) 5.7 (3.4) 6.8 (5.3) 5.5 (2.5) 6.5 (5.0)

Iron (mg) 11.2 (9.4) 10.9 (11.4) 10.3 (7.2) 11.7 (11.8)

198

Vitamin A (RE) 559 (608) 519 (807) 588 (708) 514 (650)

Vitamin D (μg) 0.9 (1.6) 0.7 (1.2) 0.8 (1.7) 0.8 (1.2)

Ascorbic acid (mg) 121 (112) 125 (150) 123 (126) 121 (125)

Folate (μg) 351 (227) 298 (270) 330 (225) 338 (256) ap-value <0.001, bp-value <0.05. 1Characteristic values are means ± standard deviation (SD). FFMI = fat-free mass index, FMI = fat mass index, BMI = body mass index, W = women, and M = men. CHO = carbohydrate, Fat-free mass wasting = FFMI <16.7 kg/m2 for men and <14.6 kg/m2 for women, BMI wasting = <18.5 kg/m2 for men and women.

199

Table 7:8 Dietary intake of 24-hour recall among tuberculosis patients with/without severe clinical TBscore (>5)

Severe TBscore Characteristic Mild ≤5 Moderate/severe >5 [mean, (SD)]1 (n=39) (n=24)

Energy (kcal) 2215 (940) 1460 (722)b

Protein (g) 70.5 (46.4) 40 (20.6)b

Total fat (g) 61.6 (50.3) 35.2 (30.6)b

Carbohydrate (g) 348 (130) 252 (128)b

Dietary fiber (g) 42.4 (27.4) 34.5 (23.4)

Protein, % energy 12.5 (7.6) 11.3 (4.6)

Fat, % energy 21.8 (10.8) 19.1 (10.3)

CHO, % energy 65.7 (13.4) 69.7 (11.8)

Calcium (mg) 772 (1330) 308 (344)b

Magnesium (mg) 312 (245) 220 (177)

Zinc (mg) 7.2 (5.0) 5.3 (3.3)

Iron (mg) 12.5 (10.9) 10.2 (9.5)

200

Vitamin A (RE) 751 (798) 421 (555)b

Vitamin D (μg) 0.7 (1.2) 0.9 (1.6)

Ascorbic acid (mg) 151 (131) 105 (119)

Folate (μg) 452 (283) 262 (179)b

ap-value <0.001, bp-value <0.05. 1Characteristic values are means ± standard deviation (SD). FFMI = fat-free mass index, FMI = fat mass index, CHO = carbohydrate, BMI = body mass index, W = women, and M = men. Fat-free mass wasting = FFMI <16.7 kg/m2 for men and <14.6 kg/m2 for women, BMI wasting = <18.5 kg/m2 for men and women.

201

CHAPTER 8

CORRELATES OF DIETARY INTAKE AMONG HIV POSITIVE AND HIV

NEGATIVE ADULTS WITH OR WITHOUT TUBERCULOSIS IN URBAN

KAMPALA, UGANDA

202

Abstract

Background Although understanding nutritional intake is important in the management of tuberculosis, its assessment in clinical and in research practice is often overlooked.

Objective We sought to determine correlates of energy and protein intake, and correlates

of inadequate dietary intake.

Methods In a cross-sectional study of 131 HIV positive and HIV negative adults with or

without tuberculosis who were enrolled from urban Uganda; 24-hour dietary intake recall

was assessed.

Results There was female gender interaction between having tuberculosis and reduced

appetite, and between having tuberculosis and being a current alcohol taker in the model

for energy intake. Women that had tuberculosis with reduced appetite or tuberculosis with history of taking alcohol had decreased energy intake. Also women who had history of alcohol intake had decreased protein intake. There was no compromise with energy and protein intake among men. Women were associated with inadequate iron intake.

Further, women with tuberculosis were associated with inadequate folate intake.

Individuals with tuberculosis residing in households of more than two people or those with no or low education were associated with inadequate vitamin A intake.

Conclusions The present study revealed that correlates of energy and protein intake

differ by gender. Women and individuals having tuberculosis who reside in overcrowded

households or who have no or low education are at vulnerable state of inadequate nutrient

intake. Further studies are needed to evaluate changes in nutrient intake and the impact

on survival.

203

Background

In sub-Saharan Africa, tuberculosis is the most common cause of death from a curable

infectious disease (Frieden et al. 2003). The incidence of clinical tuberculosis has

increased in sub-Saharan Africa due to the human immunodeficiency virus

(HIV/acquired immune-deficiency syndrome (AIDS) pandemic (Harries et al. 2001;

Lawn and Churchyard 2009), and despite adequate chemotherapy for tuberculosis,

mortality is still high among those co-infected with HIV (Haller et al. 1999).

Tuberculosis and HIV infections are both independently associated with body wasting and malnutrition. The wasting may be caused by a combination of decreased appetite, leading to a decrease in energy intake, interacting with increased losses and altered metabolism as part of the inflammatory and immune responses (Paton et al. 1999; Paton et al. 2003; Macallan et al. 1998). Moreover, wasting leads to impaired physical function

(Harries et al. 1988) and increased mortality in patients with tuberculosis (Mehta J.B et al. 1996; Rao et al. 1998; Zachariah et al. 2002; Mitnick et al. 2003). Although body wasting and malnutrition appear to separately play an important role in the clinical course of patients with tuberculosis and HIV and among those with co-infection, nutritional status and nutritional intake are often overlooked in clinical practice and in tuberculosis programs. Dietary intake studies are limited to characterize the nutritional intake, nutritional status and outcomes among patients with tuberculosis and HIV infection.

204

In the present cross-sectional study, we sought to determine correlates of energy and

protein intake, and correlates of inadequate dietary intake in a population of HIV positive

and HIV negative patients with or without tuberculosis enrolled from urban Kampala,

Uganda.

Methods

In a cross-sectional study, we enrolled 132 participants age 18 years or older residing in

Kampala district or 20 km from the study site if residence was outside Kampala in

Uganda. Data collection was conducted between November 2007 and March 2008, a period that coincides with harvesting and light rains in November and December and dry season in January and February. One participant was excluded from the analysis because of prior tuberculosis treatment. The study was conducted at the National Tuberculosis and Leprosy Program (NTLP) Clinic of the national tertiary teaching hospital, Mulago complex. Of the 131 participants who were included in the analysis, 31 were HIV positive and 32 were HIV negative with tuberculosis and were recruited at the Mulago

NTLP Clinic; 38 were HIV positive patients without tuberculosis and were recruited at the Infectious Disease Institute Clinic (IDI) located 500 meters from the Mulago NTLP

Clinic; and 30 were HIV negative individuals without tuberculosis and were recruited from the community where enrolled tuberculosis patients resided. The institutional review boards at Case Western Reserve University and Joint Clinical Research Center approved the study, with final approval by the Uganda National Council for Science and

Technology. All participants provided written informed consent to the study.

205

All subjects in the study were given appropriate pre- and post-test HIV counseling and

AIDS education. HIV-1 infection was diagnosed on the basis of a positive enzyme-linked immunosorbent assay for HIV-1 antibodies (Recombigen; Cambridge Biotech,

The dietary intake assessment was made using a single 24-hour dietary recall with open ended questions. The reference period for the 24-hour recall was the day prior to the day of the interview. In all instances, the interview was held only with the interviewee, no one

else was present except for children. The questionnaire was pre-tested by administering it

to 8 individuals selected randomly from the neighboring community to the study site. The

assessment was conducted by four trained study nursing staffs and continuously

supervised by a nutritionist using local food photographs, portion-size images, and

volumetric vessels to increase the accuracy of the recall from the previous 24 hours. The

nutritive value of raw ingredients was computed using the East African food composition

table database whose database was imported into the NutriSurvey software

(http:www.nutrisurvey.de) to easy the computations. The database used was

predominantly for local Ugandan diet. When the East African food composition table was

206

found deficient in certain food items, the United States Department of Agriculture database and the African composition table were used.

To estimate the nutrient adequacy of the diet, we calculated the nutrient adequacy ratio

(NAR) (%) for 11 micronutrients, energy, protein, fat, carbohydrate, and dietary fiber.

The NAR for a given nutrient is the ratio of a participant’s intake to the daily recommended allowance for the participant’s sex. The Food and Agriculture

Organization/World Health Organization 2002 Human Vitamin and Mineral requirements

(FAO/WHO 2002) were used for vitamin A, vitamin B6, vitamin C, vitamin D, thiamin, riboflavin, folate, magnesium, calcium, iron, and zinc whereas energy, fat, protein, carbohydrate, and dietary fiber; the Panel on Macronutrients, Panel on the Definition of

Dietary Fiber, Submcommittee on Upper Reference Levels of Nutrients, Subcommittee on Interpretation and Uses of Dietary Reference Intakes Food and Nutrition Board

(Dietary Reference Intake for Energy, Carbohydrate, Fiber, Fat, Fatty Acids, Cholesterol,

Protein, and Amino Acids (Macronutrients). A report of the Panel on Macronutrients,

Subcommittees on Upper Reference Levels of Nutrients and Interpretation and Uses of

Dietary Reference Intakes, and the Standing Committee on the Scientific Evaluation of

Dietary Reference Intakes 2005) was used. The recommendation for adequate intake was used for vitamin D (Atkinson and Ward 2001). In the case of iron and zinc, the category for moderate bioavailability was used.

207

Analytic strategy

We performed bivariate, univariate, and multivariable analyses. Chi-square and Fisher’s

Exact tests were used to compare proportions. Fisher’s Exact test was used when tabular counts were less than 5. Mann-Whitney test was used to compare dietary intake variables due to lack of normality. To determine the independent correlates of absolute energy and protein intakes, multivariable linear regression analyses including all variables that were associated with a p < 0.50 in the unadjusted analyses (Dales and Ury 1978). The following variables were evaluated in unadjusted analyses: older age group >30 years, having tuberculosis and HIV infection, no or low level (primary) of education, being separated or divorced, being single, having a household number of more than 2 people, unemployment, having no personal income, current history of alcohol intake, and reduced appetite were evaluated.

We controlled for body size by including height in the regression model for energy intake because half of the participants had tuberculosis which could in itself lead to changes in intake. Furthermore, height was correlated with energy intake. The results of the multi-

regression analyses are presented along with the mean unadjusted intakes for each

subgroup variable considered. We adjusted for energy intake in all regression analyses of

protein intake to address the composition of the participant’s diets, independent of

absolute intake. The R square was used to establish variables that were important for the

reduced models. The residuals of the final models were found to be normally distributed

for energy and protein intake. We evaluated for two-way interactions between

208

tuberculosis or between HIV and all the important variables for both the energy and

protein intake models. Significant interactions were found between tuberculosis and

alcohol intake and between tuberculosis and reduce appetite in the energy intake model

for women.

Correlates of dietary inadequacy, i.e., intakes less than the recommended daily allowance were identified with preliminary unadjusted analyses by using linear models with a log link function and binomial distribution for prevalence response variable (Wacholder

1986). All variables that showed a p-value <0.50 in the unadjusted analyses for energy,

protein, vitamins A and D, folate, and zinc intakes were included in a multivariable full

model (Mickey and Greenland 1989). The Quasi-likelihood Information Criteria (QIC)

and Quasi-likelihood under the independence Criteria (QICu) goodness fit tests were

used in choosing the best models that fit the data. The analysis was performed using SAS

GENMOD procedure (SAS Institute 2003). We evaluated for two-way interactions

between tuberculosis, between HIV, or between gender and all the important variables for

models with energy, protein, vitamins A and D, folate, and zinc deficiency as main

outcome. Significant interactions were found between female gender and reduced

appetite for the model with protein deficiency, between tuberculosis and being more than

people per household for the model with vitamin A deficiency, and between tuberculosis

and gender for the model with folate deficiency as main outcomes. The prevalence ratios

and 95% confidence intervals (CIs) for each of the independent correlates of dietary

inadequacy on the basis of binomial regression models with log link function have been

209

reported in this paper (Mickey and Greenland 1989). All analyses were performed using

SAS version 9.2 (Cary software, North Carolina SAS Institute Inc 2004).

Results

The characteristics of the study population are shown in Table 8:1. Of the 131 participants who were included in the analysis, 53% were females, 47% were males, 53% were HIV positive, and 48% had tuberculosis. Women and men differed significantly by employment and income. Women had a higher proportion of individuals who were unemployed and without income compared to men. However, men had higher a proportion of individuals who were single compared to women.

The variables that were correlated with energy and protein intakes at a significance level of p < 0.50 in the unadjusted analysis are shown in Tables 8:2 and 8:3. For each of these variables, the mean energy and protein intakes are shown. None of the bivariate energy means was significantly different among both women and men. Also the bivariate protein means were comparable among women. There were significant interactions between having tuberculosis and reduced appetite (p=0.018), and between having tuberculosis and being a current alcohol taker in the model for energy intake among women (p=0.020)

(Table 8:2). Women that had tuberculosis and had reduced appetite or were current alcohol takers had decreased energy intake. Further, women who were current alcohol takers had decreased protein intake (Table 8:3). Energy and protein intakes were not

210

affected by any factor among men (Table 8:2 and 8:3). Of note, having HIV infection did influence energy and protein intakes regardless of gender.

The correlates of inadequate intakes of energy, protein, vitamin A and D, zinc, iron, and calcium are shown in Tables 8:4, 8:5, and 8:6. Intakes were considered inadequate if they were below the recommended daily allowance. Energy intake was inadequate in 73%

(96/131) of the participants. Inadequate intakes of protein, vitamin A and D, zinc, iron, folate, and calcium were found in 60%, 74%, 98%, 56%, 86%, 63%, and 95% of participants, respectively. Women were strongly associated with inadequate protein and iron intakes whereas men were associated vitamin A inadequate intake (Table 8:4 and

8:5). Participants with no or low level of education and participants that were not married

(singles), had inadequate vitamin A and iron intake, respectively. Furthermore, participants with reduced appetite were associated with inadequate vitamin A and folate intakes.

The multivariable analyses of the correlates of inadequate energy, protein, vitamins A and D, folate, and zinc intakes are shown in Table 8:6. The variables that correlated with inadequate intake for energy, protein, vitamins A and D, folate, and zinc at p < 0.50 in the unadjusted models were evaluated for their importance in the full model are presented in

Table 8:6 together with results of the reduced models. No factor was found to influence inadequate energy intake. Being a female was significantly associated with inadequate vitamin A and iron intake, and being a female with reduced appetite was associated with

211

inadequate protein intake. Having no or low level of education and being more than two

people per household in the face of tuberculosis were significantly associated with

inadequate vitamin A intake. Further, being a female with tuberculosis and having

reduced appetite were associated with inadequate folate intake. No factor influenced

inadequate vitamin D and zinc intakes.

Discussion

We sought to determine correlates of energy and protein intake, and correlates of inadequate dietary intake in this cross-sectional study of 131 HIV positive and HIV negative adults with or without tuberculosis who were enrolled from urban Uganda.

Correlates of energy and protein dietary intakes differed by gender. Protein intake among women was influenced by alcohol intake whereas energy intake among women with tuberculosis was influenced by reduced appetite or alcohol intake. Energy intake was not affected by any study variables. Gender, reduced appetite, level of education, and tuberculosis were important factors that were associated with inadequate intakes for vitamin A, iron, protein, and folate.

The caveats to the interpretation of results in the present study are not limited to the cross-sectional nature of the study but also use of a single 24-hour dietary recall in

nutritional assessment that we cannot comment on the seasonal effects or habitual

consumption. Findings in this study suggest the vulnerability of the female gender with

compromised nutrient intake and are at worse state of inadequacy in nutrient intake

212

compared to male gender. Also individuals in overcrowded households and with no or

low level of education in the face of tuberculosis are at worse state of nutrient

inadequacy. There was female gender interaction between having tuberculosis and

reduced appetite, and between having tuberculosis and being a current alcohol taker in

the model for energy intake. Women that had tuberculosis with reduced appetite or tuberculosis with history of taking alcohol had decreased energy intake. Also women who had history of alcohol intake had decreased protein intake. There was no compromise with energy and protein intake among men. Women were associated with inadequate iron intake. Further, women with tuberculosis were associated with inadequate folate intake. Individuals with tuberculosis residing in households of more than two people or those with no or low education were associated with inadequate vitamin A intake.

The gender differences in correlates of energy and protein dietary intake in this study population suggests the cultural factors that may compromise intake among women. The vulnerability of individuals with tuberculosis residing in households of more than two people or those with no or low education to inadequate vitamin A intake reflects on the poverty level of the community. For example, unequal distribution of food within households (Carloni 1981; de Hartog A.P 1972), or men may have the opportunity to eat a wider variety or better quality foods outside the home, such as cafes or local restaurants

(Holmboe-Ottesen G and Wandel M 1991). The unequal distribution of food within households result from several factors such as women may be trained to show restraint in

213

eating, to give the best foods to men, or to allow others in the family to eat first (Lado

1992; O'Laughlin B 1974; Rosenberg E.M 1980; Dey J 1981).

The present study revealed that correlates of energy and protein intake differ by gender.

Women and individuals having tuberculosis who reside in overcrowded households or who have no or low education are at vulnerable state of inadequate nutrient intake.

Further studies are needed to evaluate changes in nutritient intake and the impact on survival.

Acknowledgements

We thank all study staff members of the Case Western Reserve University and Makerere

University research collaboration at the Tuberculosis Research Unit in the United States and in Uganda for their assistance; the faculty of staff at Case Western Reserve

University Department of Epidemiology and Biostatistics for the guidance in analyzing the project; and the Fogarty International Center, for the continued support.

This study was supported in part by the AIDS International Training Research Program,

Fogarty International Center, Grant No. TW000011, based at Case Western Reserve

University, Department of Epidemiology and Biostatistics. This work was part of Ezekiel

Mupere’s PhD thesis at Case Western Reserve University.

214

Table 8:1 Characteristics of the study population (n=131)

Characteristics All subjects Women Men p-value (n=131) (n=70) (n=61)

n (%) n (%) n (%)

Age in years

≤30 94 (72) 52 (74) 42 (69) 0.491

>30 37 (28) 18 (26) 19 (31)

HIV status

Negative 62 (47) 28 (40) 34 (56) 0.072

Positive 69 (53) 42 (60) 27 (44)

Tuberculosis

No 68 (52) 35 (50) 33 (54) 0.640

Yes 63 (48) 35 (50) 28 (46)

Education

None/primary level 69 (53) 29 (41) 33 (54) 0.147

Secondary level 62 (47) 41 (59) 28 (46)

Tribe

215

Muganda 57 (44) 30 (43) 27 (44) 0.872

Others 74 (56) 40 (57) 34 (56)

Marital status

Married 56 (43) 29 (41) 27 (44) 0.744

Single 75 (57) 41 (59) 34 (56)

Household number

One to two 40 (31) 17 (24) 23 (38) 0.100

>2 people 91 (69) 53 (76) 38 (62)

Employed

No 54 (41) 36 (51) 18 (30) 0.011

Yes 77 (59) 34 (49) 43 (70)

Income

Not at all 47 (36) 35 (50) 12 (80) <0.001

Yes 84 (64) 35 (50) 49 (20)

Takes alcohol

No 98 (75) 51 (73) 47 (77) 0.581

Yes 33 (25) 19 (27) 14 (23)

216

Appetite

No 42 (32) 27 (61) 15 (75) 0.087

Yes 89 (68) 43 (39) 46 (25)

217

Table 8:2 Correlates of dietary energy intake among women and men

Characteristics Women Multivariate Men Multivariate model (n=70) (n=61) model

Mean Estimate Mean Estimate

(SD) (SE) (SD) (SE)

Age in years

≤30 1437 (683) 2184 (849)

>30 1815 (1153) 425.0 (235.0) 1976 (729) -

HIV status

Negative 1583 (530) 2229 (775)

Positive 1544 (1010) - 1981 (854) -297.4 (234.8)

Tuberculosis

No 1630 (917) 2161 (723)

Yes 1489 (776) 698.3 (341.5)b 2070 (920) -

Education

Secondary level 1546 (950) 2121 (841)

None/primary level 1579 (689) -46.4 (209.9) 2117 (803) 110.7 (224.5)

218

Marital status

Married 1731 (783) - 2133 (863)

Single 1439 (878) 2108 (786) 126.1 (231.6)

Household number

One to two 13228 (974) 2080 (682)

>2 1634 (797) 355.1 (239.3) 2143 (892) -

Income

Yes 1637 (796) 2087 (912)

Not at all 1482 (899) 118.0 (204.5) 2127 (798) -

Takes alcohol

No 1491 (776) 2060 (813)

Yes 1744 (1013) 682.9 (298.1)b 2319 (813) 238.7 (250.5)

Appetite

Normal 1463 (989) 1862 (770)

Reduced 1621 (749) 830.3 (498.8) 2203 (818) -370.1 (250.0)

TB*reduced - -425.2 (584.0)b - - appetite

219

TB*alcohol intake -1161.0 - - - (484.5)b

R2 0.28 0.10 ap-value <0.001, bp-value <0.05; bivariate p-values. Energy intakes were adjusted for height. Multivariate models show only those variables for which p < 0.50 in the unadjusted analyses.

220

Table 8:3 Correlates of dietary protein intake among women and men

Women Multivaiate Multivariate Characteristics Men (n=61) (n=70) model model

Mean Estimate Mean Estimate

(SE) (SD) (SD) (SE)

Age in years

≤30 36.6 (23.9) 59.0 (27.3)

>30 52.1 (43.2) 3.0 (6.5) 56.1 (41.9) -

Tuberculosis

No 37.3 (21.6) 58.6 (28.4)

Yes 46.8 (36.7) 10.5 (5.4) 57.6 (36.6) -

Education

No or primary level 45.5 (25.6) 60.0 (37.4)

Secondary level 39.6 (33.3 - 56.0 (25.3) -5.2 (5.0)

Tribe

Muganda 37.8 (22.7) 60.9 (35.5)

Others 45.3 (34.8) -8.2 (5.5) 56.0 (29.7) -

221

Marital status

Married 45.2 (25.8) 61.0 (39.8)

Single 39.9 (33.2) - 55.9 (25.0) 3.4 (5.3)

Household number

One to two 42.7 (47.1) 54.3 (18.7)

>2 41.9 (23.1) - 60.4 (38.2) 4.6 (5.6)

Takes alcohol

No 44.7 (34.0) 58.3 (35.2)

Yes 35.1 (15.0) -13.9 (6.2)b 57.5 (20.0) -

Appetite

No 43.7 (34.1) 61.5 (31.2)b

Yes 39.4 (23.4) - 47.7 (33.9) -4.1 (6.0)

R2 0.50 0.67

ap-value <0.001, appetite (p=0.036) bp-value <0.05; bivariate p-values. Protein intakes were adjusted for energy intakes. Multivariate models show only those variables for which p < 0.50 in the unadjusted analyses.

222

Table 8:4 Proportions of adult individuals with inadequate dietary intakes of energy, protein, and micronutrients in relation to socio-demographic, HIV, and tuberculosis variables in Kampala, Uganda

Vitamins Characteristics Energy Protein A D

Gender

Female 50/70 (71) 48/70 (69)b 46/70 (66)b 69/70 (99)

Male 46/61 (75) 30/61 (49) 51/61 (84) 60/61 (98)

Age in years

≤30 69/94 (73) 55/94 (59) 69/94 (73) 93/94 (99)

>30 27/37 (73) 23/37 (62) 28/37 (76) 36/37 (97)

HIV status

Negative 43/62 (62) 36/62 (58) 47/62 (76) 62/62 (100)

Positive 53/69 (77) 42/69 (61) 50/69 (72) 67/69 (97)

Tuberculosis

No 50/68 (74) 38/68 (56) 54/68 (79) 68/68 (100)

Yes 46/63 (73) 40/63 (63) 43/63 (68) 61/63 (97)

223

Education

Secondary level 43/62 (69) 35/62 (56) 41/62 (66)b 61/62 (98)

None/primary 43/69 (62) 56/69 (81) 53/69 (77) 68/69 (99) level

Tribe

Muganda 41/57 (72) 35/57 (61) 45/57 (79) 56/57 (98)

Others 55/74 (74) 43/74 (58) 52/74 (70) 73/74 (99)

Marital status

Married 39/56 (70) 33/56 (59) 36/56 (64)b 55/56 (98)

Single 57/75 (76) 45/75 (60) 61/75 (81) 74/75 (99)

Household number

One to two 32/40 (80) 55/91 (60) 60/91 (66) 90/91 (99)

>2 64/91 (70) 23/40 (58) 37/40 (93) 39/40 (98)

Employed

Yes 59/77 (77) 47/77 (61) 58/77 (75) 75/77 (97)

No 37/54 (69) 31/54 (77) 39/54 (72) 54/54 (100)

Income

224

Yes 64/84 (76) 51/84 (61) 63/84 (75) 82/84 (98)

Not at all 32/47 (68) 27/47 (57) 34/47 (72) 47/47 (100)

Takes alcohol

No 73/98 (74) 57/98 (58) 71/98 (72) 96/98 (98)

Yes 23/33 (70) 21/33 (64) 26/33 (79) 33/33 (100)

Appetite

Normal 62/89 (70) 48/89 (54) 71/89 (80)b 87/89 (98)

Reduced 30/42 (71) 26/42 (62) 42/42 34/42 (81) (1000 ap-value <0.001, bp-value <0.05; bivariate p-values.

225

Table 8:5 Proportions of adult individuals with inadequate dietary intakes of energy, protein, and micronutrients in relation to socio-demographic, HIV, and tuberculosis variables in Kampala, Uganda

Minerals Characteristics Zinc Iron Folate Calcium

Gender

Female 41/70 (59) 66/70 (94)b 46/70 (66) 67/70 (96)

Male 33/61 (54) 47/61 (77) 36/61 (59) 57/61 (93)

Age in years

≤30 51/94 (54) 81/94 (86) 56/94 (60) 90/94 (96)

>30 23/37 (62) 32/37 (86) 26/37 (70) 34/37 (92)

HIV status

Negative 36/62 (58) 53/62 (85) 34/62 (55) 60/62 (97)

Positive 38/69 (55) 60/69 (87) 48/69 (70) 64/69 (93)

Tuberculosis

No 38/68 (56) 58/68 (85) 39/68 (57) 65/68 (97)

Yes 36/63 (57) 55/63 (87) 43/63 (68) 59/63 (59)

226

Education

Secondary level 37/62 (60) 51/62 (82) 40/62 (65) 59/62 (95)

None/primary level 37/69 (54) 62/69 (90) 42/69 (61) 65/69 (94)

Tribe

Muganda 32/57 (56) 48/57 (84) 38/57 (67) 53/57 (93)

Others 42/74 (57) 65/74 (88) 44/74 (59) 71/74 (96)

Marital status

Married 29/56 (52) 50/56 (89) 35/56 (63) 53/56 (95)

Single 45/75 (60) 63/75 (84) 47/75 (63) 71/75 (95)

Household number

One to two 52/91 (57) 81/91 (89) 58/91 (64) 85/91 (93)

>2 22/40 (55) 32/40 (80) 24/40 (60) 39/40 (98)

Employed

Yes 44/77 (57) 65/77 (84) 49/77 (49) 73/77 (95)

No 30/54 (56) 48/54 (88) 33/54 (61) 51/54 (94)

Income

Yes 45/84 (54) 69/84 (82) 56/84 (67) 78/84 (93)

227

Not at all 29/47 (62) 44/47 (94) 26/47 (55) 46/47 (98)

Takes alcohol

No 56/98 (57) 87/98 (89) 63/98 (64) 94/98 (96)

Yes 18/33 (55) 26/33 (79) 19/33 (58) 30/33 (91)

Appetite

Normal 48/89 85/89 (96) 49/89 (55) 76/89 (85) (54)b

Reduced 25/42 (60) 37/42 (88) 34/42 (81) 39/42 (93) ap-value <0.001, bp-value <0.05; bivariate p-values.

228

Table 8:6 Correlates of inadequate dietary intakes of key nutrients from multivariate models among adult individuals in Kampala, Uganda

Characteristics Full model1 Reduced model2

Prevalence ratio p- Prevalence ratio p-

(95% CI) value (95% CI) value

Inadequate energy intake

HIV positive 1.07 (0.85, 1.34) 0.58 1.07 (0.85, 1.34) 0.58

No education/primary level 1.08 (0.87, 1.34) 0.62 1.08 (0.87, 1.34) 0.62

Single 0.91 (0.73, 1.14) 0.42 0.91 (0.73, 1.14) 0.42

>2 per household 0.91 (0.73, 1.14) 0.43 0.91 (0.73, 1.14) 0.43

No income 0.93 (0.73, 1.17) 0.52 0.93 (0.73, 1.17) 0.52

Reduced appetite 1.16 (0.94, 1.43) 0.16 1.16 (0.94, 1.43) 0.16

QIC 900.8 900.8

Inadequate protein intake

Female 1.34 (0.99, 1.81) 0.06 1.62 (1.08, 2.43) 0.020

Tuberculosis 0.97 (0.68, 1.40) 0.88 0.97 (0.68, 1.40) 0.88

No education/primary level 1.05 (0.79, 1.40) 0.73 1.06 (0.80, 1.40) 0.69

229

Single 1.29 (0.90, 1.85) 0.17 0.98 (0.74, 1.30) 0.87

Reduced appetite 0.96 (0.72, 1.28) 0.79 1.81 (1.81, 3.02) 0.023

Female*reduced appetite - - 0.59 (0.34, 1.03) 0.063

QIC 564.1 554.6

Inadequate Vitamin A

Female 0.80 (0.66, 0.98) 0.029 0.75 (0.61, 0.92) 0.01

Tuberculosis 0.95 (0.76, 1.18) 0.62 1.35 (1.04, 1.74) 0.02

No education/primary level 1.28 (1.05, 1.57) 0.015 1.28 (1.05, 1.56) 0.01

Other than Muganda 1.19 (0.98, 1.44) 0.08 1.16 (0.96, 1.40) 0.14

Single 0.84 (0.68, 1.04) 0.11 0.83 (0.68, 1.02) 0.08

>2 per household 0.77 (0.65, 0.92) 0.004 1.03 (0.79, 1.34) 0.84

Takes alcohol 1.00 (0.82, 1.24) 0.96 1.06 (0.87, 1.29) 0.57

Reduced appetite 0.83 (0.64, 1.09) 0.17 0.85 (0.66, 1.10) 0.23

TB*>2 per household - - 0.56 (0.39, 0.82) 0.003

QIC 929.8 919.0

Inadequate folate

Female 1.12 (0.85, 1.48) 0.42 1.57 (1.02, 2.42) 0.04

230

Age >30 years 1.13 (0.87, 1.47) 0.35 1.21 (0.91, 1.62) 0.19

HIV positive 1.17 (0.87, 1.58) 0.30 1.09 (0.83, 1.43) 0.52

Tuberculosis 0.99 (0.70, 1.39) 0.94 0.82 (0.64, 1.06) 0.13

No education/primary level 0.84 (0.65, 1.09) 0.19 0.99 (0.77, 1.27) 0.94

Other than Muganda 1.02 (0.79, 1.34) 0.83 0.73 (0.50, 1.09) 0.12

Single 0.94 (0.72, 1.22) 0.65 0.97 (0.75, 1.26) 0.81

No income 0.82 (0.60, 1.12) 0.21 0.83 (0.61, 1.13) 0.23

Reduced appetite 1.45 (1.03, 2.03) 0.03 1.48 (1.07, 2.04) 0.02

TB*female - - 1.92 (1.13, 3.28) 0.02

QIC 603.1 600.8

Iron

Female 1.20 (1.03, 1.40) 0.02 1.20 (1.03, 1.40) 0.02

Single 1.05 (0.91, 1.22) 0.49 1.05 (0.91, 1.22) 0.49

>2 people per household 1.03 (0.85, 1.25) 0.74 1.03 (0.85, 1.25) 0.74

Unemployment 0.91 (0.73, 1.13) 0.40 0.91 (0.73, 1.13) 0.40

No income 1.16 (0.93, 1.44) 0.20 1.16 (0.93, 1.44) 0.20

Takes alcohol 0.89 (0.75, 1.06) 0.19 0.89 (0.75, 1.06) 0.19

231

QIC 1800.4 1800.4

Vitamin D

HIV positive 0.95 (0.90, 1.01) 0.13 0.95 (0.90, 1.01) 0.13

Tuberculosis 0.92 (0.83, 1.03) 0.13 0.92 (0.83, 1.03) 0.13

No education/primary level 1.01 (0.97, 1.06) 0.57 1.01 (0.97, 1.06) 0.57

Unemployment 1.00 (0.98, 1.03) 0.81 1.00 (0.98, 1.03) 0.81

No income 1.03 (0.99, 1.08) 0.65 1.03 (0.99, 1.08) 0.65

Takes alcohol 1.02 (0.99, 1.04) 0.22 1.02 (0.99, 1.04) 0.22

Reduced appetite 1.09 (0.98, 1.22) 0.13 1.09 (0.98, 1.22) 0.13

QIC 16188.0 16188.0

Inadequate zinc intake

Age >30 years 1.24 (0.89, 1.72) 0.21 1.24 (0.89, 1.72) 0.21

No education/primary level 0.88 (0.65, 1.20) 0.43 0.88 (0.65, 1.20) 0.43

Single 0.86 (0.63, 1.17) 0.33 0.86 (0.63, 1.17) 0.33

No income 1.16 (0.86, 1.58) 0.33 1.16 (0.86, 1.58) 0.33

QIC 520.7 - 520.7 -

1The adjusted analyses first controlled for all variables that showed an association with inadequate intake of the outcome variable at p < 0.50 in unadjusted models. 2Shows best model.

232

CHAPTER 9

IMPACT OF BODY WASTING ON SURVIVAL AMONG ADULT PATIENTS

WITH PULMONARY TUBERCULOSIS IN URBAN KAMPALA, UGANDA

233

Abstract

Background Body mass index (BMI) may over estimate body wasting and mortality due to wasting because it is insensitive to body fatness at low BMI and above normal muscle development. We assessed the impact of HIV and body wasting on survival in tuberculosis patients using precise measures of nutritional status, the height-normalized fat-free mass (FFMI) and fat mass (FMI) indices.

Methods In a retrospective cohort study of adult patients, 310 patients with baseline wasting and 437 without wasting as measured by BMI (kg/m2); 103 with baseline

wasting and 208 without wasting as measured by fat-free mass index (kg/m2); and 401

HIV positive and 346 HIV negative patients were followed for survival.

Results During the follow-up period, 19% of 310 patients with baseline wasting by BMI

died compared to 11% of 437 without wasting, a crude risk ratio of 1.74 (95% confidence

interval (CI): 1.22, 2.48). Of 103 with baseline wasting by fat-free mass index, 16% died,

compared to 7% without wasting, crude risk ratio of 2.31 (95% CI: 1.10, 3.92). In

stratified survival analysis, survival proportion was significantly lower among men with

reduced BMI compared to men with normal BMI; and lower among women with reduced

fat-free mass index compared to women with normal fat-free mass index. In multivariable Cox regression model using anthropometric data, the relative hazard of death when patient had reduced BMI was 1.85 (95% CI: 1.25, 2.73). In a nested model, the relative hazard for death was 1.70 (95% CI: 1.03, 2.81) for men with reduced BMI and 1.83 (95% CI: 0.96, 3.50) for women with reduced BMI. In a model using fat-free

234

mass index data, the relative hazard of death when patients had reduced fat-free mass index was 1.88 (0.96, 3.65). In a nested model, the relative hazard of death was 6.83

(95% CI: 2.14, 21.74) for women with reduced fat-free mass index compared to women with normal fat-free mass and 0.80 (95% CI: 0.35, 1.84) for men with reduced fat-free mass index. In Kaplan-Meier analysis, men had significantly lower survival compared to women (p=0.016) Cox regression analysis HIV positive men had 1.62 (95% confidence interval (CI): 1.05, 2.52) hazard of death compared to HIV positive women. HIV negative men had 0.57 (95% CI: 0.10, 3.31) hazard of death compared to HIV negative women.

Conclusion Findings show that body wasting is associated with reduced survival that differed by gender. BMI is a better predictor of death among men whereas fat-free mass index among women. There was gender differences in survival with poor outcomes among HIV positive men.

235

Background

Tuberculosis is among the foremost infectious cause of mortality worldwide. Globally,

recent reports showed 9.2 million new cases and 1.7 million tuberculosis-associated

deaths to have occurred in 2006 (Vitoria et al. 2009). In HIV-infected individuals, tuberculosis is the leading cause of death accounting for up to 11% of AIDS-related mortality worldwide (Corbett et al. 2003; Holmes et al. 2003). In sub-Saharan Africa

where there is a high burden of both TB and human immunodeficiency virus (HIV)

infection, case fatality rates for tuberculosis in HIV-infected patients are extremely high

(up to 40%) (Corbett et al. 2003; Mugusi et al. 2009). Tuberculosis is frequently found at

autopsy of acquired immunodeficiency syndrome (AIDS) patients (Lucas et al. 1993),

particularly cachectic patients suggesting that tuberculosis exacerbates the wasting

process of HIV-infected people in Africa.

Body wasting is regarded as a cardinal feature of tuberculosis. A significant proportion of

African tuberculosis patients have a marked degree of wasting by the time they present

for registration and treatment (Kennedy et al. 1996; Harries et al. 1988; Zachariah et al.

2002). Wasting associated with tuberculosis is likely caused by a combination of

decreased appetite, leading to a decrease in energy intake, interacting with increased

losses and altered metabolism resulting from the inflammatory and immune responses

(Paton et al. 1999; Paton et al. 2003; Macallan 1999). Wasting is associated with

impaired physical function (Harries et al. 1988), longer hospitalization days and

236

increased mortality in patients with tuberculosis (Rao et al. 1998; Zachariah et al. 2002;

Mitnick et al. 2003). It has been reported that in patients with both TB and HIV co-

infection, the wasting process is exacerbated (Macallan 1999; Lucas et al. 1994). In

contrast, findings from several cross-sectional studies appear to show no large differences

in body composition between HIV-infected adults with tuberculosis and HIV-negative

adults with tuberculosis at presentation (Shah et al. 2001; Niyongabo et al. 1999;

Niyongabo et al. 1994; Mupere et al. 2010) suggesting that TB is the dominant factor

inducing wasting. This has been shown in several reports (Mupere et al. 2010; Paton and

Ng 2006). However, gender differences in body composition at presentation among TB

patients have been reported (Kennedy et al. 1996; Mupere et al. 2010).

Although the association of malnutrition with tuberculosis, and its impact on co-infection

with tuberculosis and HIV has been described (Macallan 1999; van Lettow, Fawzi, and

Semba 2003) in clinical settings and in several studies that have evaluated mortality in

tuberculosis (Zachariah et al. 2002; Mugusi et al. 2009), the extent of wasting has been

assessed using body mass index (BMI). However, BMI is insensitive to body fatness,

particularly at low BMI, as well as with above-normal muscle development (Kyle,

Genton, and Pichard 2002; Kyle, Piccoli, and Pichard 2003). Thus, previous estimates of

mortality between malnourished and individuals with normal nutrition might have been

overestimated. Furthermore, the confounding effects of gender and HIV infection on

survival among tuberculosis patients using precise estimates of body composition have

not yet been described. Fat and fat-free mass body composition measurements have been shown to permit a more precise evaluation of nutritional status (VanItallie et al. 1990;

237

Kyle, Piccoli, and Pichard 2003). Bioelectrical impedance analysis (BIA) has been

recommended as the practical and precise method for clinical assessment of fat and fat-

free mass (Kyle et al. 2004; Kyle, Genton, and Pichard 2002). In this study, we evaluated

the impact of body wasting using fat and fat-free mass and BMI as measures of body wasting; and the impact of HIV infection on survival among patients with pulmonary tuberculosis. Our results show for the first time gender differences of observed survival using fat-free mass.

Methodology

Study Design

We conducted a retrospective cohort study that consisted of 753 adult pulmonary tuberculosis patients having confirmed HIV status and defined baseline body wasting or not using the completed five year Household Contact (HHC) study, the completed phase

II prednisolone double blind randomized placebo controlled clinical trial, and the ongoing

Kawempe Community Health (KCH) study. Of the 753 patients, 314 were enrolled into

the HHC, 344 into the KCH, and 95 into the placebo arm of the prednisolone clinical

trial. The HHC and KCH studies were observational epidemiologic studies; organized

and conducted by the Makerere University and Case Western Reserve University

tuberculosis research collaboration (Uganda-CWRU) that has been ongoing for the last

20 years in Uganda. The HHC was the initial household contact study from 1995 to 1999 that described the epidemiology of tuberculosis in urban Kampala, Uganda (Guwatudde et al. 2003). The subsequent KCH started in 2002 and is still ongoing (Stein et al. 2005).

238

The KCH was developed specifically to focus on the determinants of host factors

associated with primary infection, re-infection, reactivation, and progression of clinical disease and to identify and track individual strains of mycobacterial tuberculosis through

Ugandan households and local community. The phase II clinical trial was conducted

between 1995 to 2000 to determine whether immunoadjuvant prednisolone therapy in

HIV-infected patients with tuberculosis who have CD4(+) T cell counts >/=200 cells/ mu

L is safe and effective at increasing CD4(+) T cell counts.

The institutional review boards at Case Western Reserve University in the United States

and Joint Clinical Research Center in Uganda reviewed the protocol and final approval

was obtained from the Uganda National Council for Science and Technology. All

patients in the HHC and KCH had written informed consent to be enrolled in the study.

All participants in both HHC and KCH were given appropriate pre- and post-test HIV counseling and AIDS education. HIV-1 infection was diagnosed on the basis of a positive enzyme-linked immunosorbent assay for HIV-1 antibodies (Recombigen; Cambridge

Biotech, Cambridge, MA). HIV-seropositive participants who were newly identified with

HIV were not on antiretroviral therapy at the time of measurement; no patients with pre- existing HIV infection at the time of household evaluation were on antiretroviral therapy because patients were enrolled before the advent of antiretroviral therapy in Uganda.

During follow-up, patients who became eligible were started on antiretroviral therapy.

239

At enrollment, basic demographic information and a medical history were collected, and

a standardized physical examination was conducted by a medical officer. Active

tuberculosis was confirmed by sputum smear microscopy and culture. Patients with

active tuberculosis were treated with standard four-drug chemotherapy for tuberculosis per guidelines of the Ugandan Ministry of Health. Adults with a previous history of treated pulmonary tuberculosis were excluded in the study. Of the 753 participants who were enrolled in the three studies, 6 were excluded because of being below 18 years of age, leaving 747 participants in total available for anthropometric analysis.

The datasets for the three studies were first tested for differences in baseline characteristics before combining for analysis. The datasets were different regarding extent of tuberculosis disease on chest x-ray because the phase II prednisolone trial enrolled only HIV-associated tuberculosis patients with CD4 cell count >200 cells/l compared to HHC and KCH studies (Appendix, Table 13:1). One of our interests was to establish the confounding effect of HIV; therefore, we combined all the three datasets for analysis. The BIA data (specifically, fat-free mass and fat mass) were collected during the KCH study only. Of the 344 participants who were enrolled in KCH, 33 were excluded due to lack of BIA measurements (n=29) data and being below 18 years of age

(n=4), leaving 311 participants in total available for BIA analysis. However, there were no differences in baseline age, gender, weight, height, BMI, smoking status, hemoglobin, chest x-ray disease extent, and history of weight loss between participants who were included and those who were excluded.

240

Measurements

In all the three studies, socio-demographic and clinical information was obtained through

standardized interviews and physical examination performed by trained medical officers.

Venous blood was collected for HIV-1 enzyme immunoassay testing and complete blood

and differential counts. HIV infection was documented by enzyme-linked immunosorbent

assays. All participants had posterior-anterior chest X-rays taken at baseline.

Expectorated sputum specimens were collected, concentrated, and stained for acid fast bacilli (AFB) with Ziehl-Neelsen stain at the Wandegeya national reference laboratory in

Uganda. AFB smears were reviewed by trained technicians who graded the smears by the number of acid-fast organisms seen on the light microscopy according to criteria established by the WHO (International Union Against Tuberculosis and Lung Disease

1986). Specimens were cultured for mycobacteria tuberculosis on Lowenstein-Jensen

medium slants, incubated at 370C in air and examined weekly until positive or for 8

weeks.

Nutritional status was assessed using anthropometric measurements such as height and

weight and BIA Detroit, MI, RJL Systems. Body weight was determined to the nearest

performed by one trained observer using the same equipment and recommended standard

conditions with regard to body position, previous exercise, dietary intake, skin

241

temperature, and voiding of the bladder were taken into consideration in taking BIA

measurements (Kyle et al. 2004). All BIA measurements during the KCH study were

performed on the day patients were confirmed to have tuberculosis disease.

The BIA is a simple, easy, safe, non-invasive technique, that has been recommended for nutritional studies in the clinical setting (Kyle et al. 2004; Kyle, Piccoli, and Pichard

1996) and have been applied elsewhere in African studies (Villamor et al. 2006; Shah et al. 2001; Van Lettow et al. 2004). Fat mass was calculated as body weight minus fat-free

mass.

Exposure variable

Baseline body wasting was the main exposure. We used BMI and height-normalized indices (adjusted for height2) of body composition that partition BMI into fat-free mass

242

index (FFMI) and fat mass index (FMI) (Schutz, Kyle, and Pichard 2002; VanItallie et al.

1990; Kyle, Piccoli, and Pichard 2003) to establish the body wasting status of

participants. The FFMI and FMI have the advantages of compensating for differences in

height and age (Kyle, Genton, and Pichard 2002). Also, the use of the FFMI and FMI

eliminates some of the differences between population groups. We defined body wasting

as patients having the low fat-free mass index (FFMI) and the low body fat mass index

(FMI) corresponding to WHO BMI categories for malnutrition as previously reported

(Table 9:1) (Kyle, Piccoli, and Pichard 2003). The FFMI <16.7 (kg/m2) for men and

<14.6 (kg/m2) for women and the FMI <1.8 (kg/m2) for men and <3.9 (kg/m2) for women

corresponds to a BMI of <18.5 kg/m2, the WHO cutoff for malnutrition (WHO Tech Rep

1995) among adults.

Table 9:1 Definitions of low and normal fat and fat-free mass index values for corresponding body mass index in adults

Characteristic Low Normal

Body mass index (BMI)a

Women and men in kg/m2 < 18.5 ≥ 18.5

Fat-free mass index (FFMI)b

Women in kg/m2 < 16.7 ≥ 16.7

Men in kg/m2 < 14.6 ≥ 14.6

243

Fat mass index (FMI)b

Women in kg/m2 < 1.8 ≥ 1.8

Men in kg/m2 < 3.9 ≥ 3.9

aWorld Health Organization categories, sex independent (WHO Tech Rep 1995). bKyle et al (Kyle, Piccoli, and Pichard 2003; Kyle et al. 2003).

Study outcome variable

Observed survival was the main study outcome. We defined observed survival as the

period between enrollment in the study and death or censoring. Participants for the three

studies were censored at the last clinic visit when they were known to be alive or at the

end of the study observation. During the conduct of the three studies at the Makerere

University-Case Western Reserve University research collaboration, mortality was assessed through a standard interview of family members or review of hospital records.

When a participant failed to keep a scheduled visit, the health visitor made a visit to the participant’s home to determine the vital status of the study participant. The family members also provided the date of death and prominent symptoms at the time of death.

Statistical analysis

Baseline characteristics for both participants with and participants without baseline wasting were compared using the χ2 test or Fisher’s exact test (where tabular counts were

less than 5) for binary data and student’s t-test for continuous variables or Wilcoxon-

244

Mann Whitney test for variables not normally distributed. Concordance in estimating wasting between the BMI cut-off of <18.5 kg/m2 and the FFMI cut-off of <16.7 (kg/m2) for men and <14.6 (kg/m2) for women, and the concordance in estimating wasting between BMI cut-off level and the FMI cut-off of <1.8 (kg/m2) for men and <3.9 (kg/m2) for women was assessed using kappa, к. Where к = 0 represents no agreement, 0.- 0.2 represents slight agreement, 0.2 – 0.4 represents fair agreement, 0.4 – 0.6 represents moderate agreement, 0.6 – 0.8 represents substantial agreement, and 0.8 – 1.0 represents almost perfect agreement, beyond chance (McGinn et al. 2004).

Mortality rates were estimated and stratified according to sex, young and old age group,

HIV status, anemia, smoking status, alcohol intake status, history of weight loss, and presence or absence of moderate/ or far advanced disease extent on chest X-ray. The incident rate of death was estimated using person-years method for the study population.

The overall distributions of survival for participants presenting with or without body wasting were estimated using the Kaplan-Meier method and compared using the log-rank test (Kaplan E.I and Meier P 1958). Overall the observed range of the data, the probabilities of death did not reach 50% so the median time from diagnosis of tuberculosis with body wasting as assessed by reduced BMI and FFMI to death could not be estimated.

A series of Cox proportional hazards models (Hosmer D.W.Jr and Lemeshow S 1999) were fit after testing for the proportional hazards assumptions for each variable. The

245

proportional hazards assumption was tested by graphical methods and goodness of fit

Schoenfeld residuals (Appendix, Table 13:2). Observed survival was the depended

variable in each model. The independent variables tested included age, sex, HIV status,

presence or absence of anemia, smoking and alcohol intake status, presence or absence of

history of weight loss, and presence or absence of moderate/or far advanced tuberculosis

disease extent on chest X-ray. Variables that were associated with survival in a univariate

analysis with p < 0.30 or with biological plausibility were evaluated in a series of multivariable models.

All two-way interactions between baseline body wasting status as measured by reduced

BMI or FFMI and the main effects were compared; the only significant interaction was with sex in the models involving FFMI variable for body wasting status. Furthermore, in our previous study (Mupere et al. 2010) we found that there are gender differences in body composition among patients with pulmonary tuberculosis. Thus, we developed separate stratified Cox models for participants according to sex status. Using a chunk wise test, interactions with sex, HIV, and age were found to be important in the model for reduced BMI whereas sex, age, and hemoglobin were important in the model for reduced

FFMI. Hence separate stratified Cox regression models were performed according to HIV status and according to young (≤30 years) or older (>30 years) age groups to understand the independent effect of HIV and age on hazard of death. All variables fulfilled the proportional hazard assumption within the gender, HIV status, and age group strata.

Models were compared using -2 log likelihood tests. All analyses were performed using

SAS version 9.1.3 (Cary software, North Carolina SAS Institute Inc. 2000 – 2004).

246

Results

Descriptive statistics

Of the 747 patients who were included in the anthropometric data analysis, 395 (53%) were men and 352 (47%) were women; 401 (54%) were HIV positive, 345 (46%) were

HIV negative, and 1 had unknown HIV status; and the mean age was 30.8 ± 8.9 SD years. Women had significantly higher BMI (20.3 ± 3.3 versus 18.6 ± 2.1) compared to men (Table 9:2). Of the 311 patients included in the BIA analysis, 164 (53%) were men and 147 (47%) were women; 138 (45%) were HIV positive, 172 (55%) were HIV negative, and 1 had unknown HIV status; and the mean age was 29.9 ± 9.2 SD years.

Women had significantly lower FFMI (15.8 ± 1.1 versus 16.6 ± 1.4) and significantly higher FMI (4.4 ± 2.8 versus 2.0 ± 0.8) compared to men (Table 9:2). Among men, FFMI had a strong positive correlation with BMI whereas among women, FMI had a strong positive correlation with BMI. These gender differences in body composition and correlation with BMI reflect on the high fat body content among women and the high fat- free mass among men as previously reported (Mupere et al. 2010). Furthermore, the strengths of correlation between BMI and FMI suggest that BMI is a measure of body fatness.

Among women, there was fair concordance [к = 0.34, (95% CI: 0.18, 0.50)] between

FFMI (women <14.6 kg/m2) and BMI (<18.5 kg/m2) cut-off for assessing body wasting with 34% of the women having both reduced FFMI and BMI whereas among men there was moderate concordance [к = 0.57, (95% CI: 0.45, 0.69)] with 80% of participants

247

having both reduced FFMI (for men <16.7 kg/m2)) and BMI (<18.5 kg/m2) (Table 9:3).

Fat mass index (FMI <3.9 kg/m2) and BMI cut-off for body wasting among women had substantial concordance [к = 0.61, (95% CI: 0.49, 0.73)] with 100% of participants having both reduced FFMI and BMI whereas among men, the concordance was moderate

[к = 0.52, (95% CI: 0.39, 0.65)] with 64% of participants having both reduced FMI (<1.8 kg/m2) and BMI (Table 9:3). Of the 747 patients who were included in the

anthropometric data analysis, 310 (42%) had body wasting at presentation and 437 (58%)

no wasting based on BMI cut-off of < 18.5 kg/m2 for both men and women. In the study

population of 311 participants that had BIA measurements, 103 (33%) had body wasting

at presentation and 208 (67%) no wasting using FFMI cut-off of <16.7 kg/m2 for men

and <14.6 kg/m2 for women. Whereas using FMI cut-off of <3.9 kg/m2 for women and

for <1.8 kg/m2 for men, 135 (43%) had baseline wasting and 176 (57%) had no wasting.

Baseline characteristics

Men had a higher frequency of reduced BMI and FFMI compared to women at presentation (Table 9:4). For example men had a proportion of 63% (194/310) for reduced BMI compared to 37% (116/310) among women. Men had a proportion of 80%

(82/103) of reduced FFMI compared to 20% (21/103) in women. Patients who presented with anemia (hemoglobin ≤10 mg/dl), history of prior smoking and current alcohol intake had a lower frequency of reduced BMI and FFMI. Patients who presented with moderate/far advanced disease on chest x-ray and history of weight loss had a higher frequency of reduced BMI and FFMI (Table 9:4). Of note, there were no differences in

248

body wasting at presentation between HIV positive and HIV negative tuberculosis patients regardless of whether BMI, FFMI, and FMI cut-off was used in assessing wasting (Appendix, Table 13:3). This suggests the lack of HIV impact on tuberculosis body wasting as previously reported (Mupere et al. 2010). HIV positive patients had a significantly higher frequency of older individuals (64% (200/312) versus 46%

(201/434)), anemic individuals (74% (89/121) versus 51% (173/342)), current alcohol intake (62% (174/279) versus 49% (226/466), and lower frequency of history of weight loss (51% (295/578) versus 64% (105/165)) than HIV negative patients at presentation.

Survival analysis

During the mean follow-up period of 31 ± 23 (SD) months for the anthropometric data, there were a total of 105 deaths. Of these 105 deaths, 76 occurred within the first 6 months during tuberculosis treatment, 18 occurred in the first year after completion of tuberculosis treatment at 6 months, and only 11 occurred more than one year after treatment. The overall annual mortality rate was 55.1 per 100 years of observation. The mortality rates were 393.6 per 100 person months of observation in the first 6 months,

326.9 per 100 person years of observation in the first year following treatment, and 6.0 per 100 person years of observation more than one year following treatment. Most deaths

(n=99) occurred among tuberculosis patients with HIV infection. In addition to HIV status, wasting as measured by BMI was also common (n=58). Of the 310 patients who presented with reduced BMI ≤ 18.5 kg/m2 19% died compared to 11% of the 437 patients

249

with normal BMI >18.5 kg/m2, leading to a crude risk ratio of 1.74 (95%: CI, 1.22 –

2.48) (Table 9:5).

Among patients with reduced BMI, 32% of the 167 that had HIV infection died

compared to 20% who died of the 234 that had HIV infection with normal BMI at

presentation, leading to a relative risk of 1.61 (95% CI: 1.50, 2.27) (Table 9:5). Among

HIV negative tuberculosis patients; however, the relative risk of death associated with

body wasting at presentation was 7.06 (95% CI: 0.83, 59.81) compared with patients

without wasting (Table 9:5). The relative risk of death associated with body mass wasting

at presentation among tuberculosis patients increased significantly after stratifying by

younger age group ≤30 years, hemoglobin >10 mg/dl, non-current alcohol intake status, and normal/or mild disease extent on chest x-ray (Table 9:5).

During the mean follow-up period of 26 ± 16 (SD) months for the BIA data, there were a total of 30 deaths. Of these 30 deaths, 15 occurred within the first 6 months during tuberculosis treatment, 13 occurred in the first year after completion of tuberculosis treatment at 6 months, and only 2 occurred more than one year after treatment. The overall annual mortality rate was 45.1 per 100 years of observation. The mortality rates were 141.4 per 100 person months of observation in the first 6 months, 283.3 per 100 person years of observation in the first year following treatment, and 3.3 per 100 person years of observation more than one year following treatment. Most deaths (n=29) occurred among tuberculosis patients with HIV infection.

250

In addition to HIV status, wasting as measured by FFMI was also common (n=16). Of the

103 patients who presented with reduced FFMI 16% died compared to 7% of the 208

patients with normal FFMI, leading to a crude risk ratio of 2.31 (95%: CI, 1.17 – 4.54)

(Table 9:6). Among patients with reduced FFMI, 32% of the 47 that had HIV infection

died compared to 15% who died of the 91 that had HIV infection with normal FFMI at

presentation, leading to a relative risk of 2.07 (95%: CI, 1.10 – 3.92) (Table 6). However,

co-infected patients with reduced FFMI had a higher mortality of 32% (15/47) compared

to 15% (14/91) among co-infected patients with normal FFMI, leading to a relative risk

of 2.07 (95% CI: 1.10, 3.92) (Table 9:6). The relative risk of death associated with body wasting at presentation among tuberculosis patients increased after stratifying by female sex, younger age group <30 years, anemia (hemoglobin ≤10 mg/dl), non -current smoking, and normal/or minimal disease extent on chest X-ray (Table 9:6). The relative risk of death associated with fat-free mass wasting at presentation among tuberculosis patients increased significantly after stratifying by female sex, normal/or mild extent of x-ray disease, anemia (hemoglobin ≤10 mg/dl), history of past smoking, and history of weight loss. The risk of death among women with fat-free mass wasting was dramatic with an estimate of 6.00 (95% CI: 2.14, 16.86) (Table 9:6).

Kaplan-Meier analysis

When we performed a Kaplan-Meier analysis using anthropometric data, the overall unadjusted survival for patients who presented with body wasting (BMI <18.5 kg/m2) had significantly lower survival compared to patients who presented with normal nutrition

251

(BMI ≥ 18.5 kg/m2) (p = 0.001, log-rank test). When the Kaplan-Meier analysis was stratified according to gender, survival proportion was significantly lower among men with reduced BMI <18.5 kg/m2 at presentation compared to men with normal BMI ≥ 18.5

kg/m2 at presentation (p = 0.033, log-rank test; Figure 9:1). For women with reduced

BMI at presentation, the survival proportion was lower than women with normal BMI,

but this difference was not statistically significant (p = 0.119, log-rank test; Appendix

Figure 13:1).

When we performed a Kaplan-Meier analysis using the BIA data, the overall unadjusted

survival for patients who presented with fat-free mass body wasting (FFMI <16.7 kg/m2

for men and <14.6 kg/m2 for women) had significantly lower survival compared to

patients who presented with normal nutrition (FFMI ≥16.7 kg/m2 for men and ≥14.6

kg/m2 for women) (p = 0.016, log-rank test). When the Kaplan-Meier analysis was stratified according to sex, survival proportion was significantly lower among women with reduced fat-free mass (FFMI <18.5 kg/m2) at presentation compared to women with

normal fat-free mass (FFMI ≥14.6 kg/m2) at presentation (p = 0.0004, log-rank test;

Figure 9:2). For men with reduced FFMI at presentation, the survival proportion was not

different from men who had normal FFMI (p = 0.647, log-rank test; Appendix Figure

13:2).

When we performed a Kaplan-Meier analysis according to sex status, men had

significantly lower unadjusted survival compared to women (p=0.035, log-rank test;

252

Appendix Figure 13:3). When the Kaplan-Meier analysis was stratified according to HIV status, survival proportion was significantly lower among HIV positive men compared to

HIV positive women (p = 0.016, log-rank test; Figure 9:3). For HIV negative men, the survival proportion was not different from HIV negative women (p = 0.937, log-rank test;

Appendix Figure 13:4).

Cox Proportional Hazard Regression

In the univariate Cox proportional hazards model, the unadjusted relative hazard for death in patients with reduced BMI <18.5 kg/m2 at presentation compared to patients who presented with normal BMI ≥18.5 kg/m2 was 1.80 (95% CI, 1.23, 2.64). Similarly, the unadjusted relative hazard for death in patients with reduced FFMI (<16.7 kg/m2 for men and <14.6 kg/m2 for women) at presentation compared to patients who presented with normal FFMI (≥16.7 kg/m2 for men and ≥14.6 kg/m2 for women) was 2.34 (95% CI,

1.14, 4.80; Table 9:7). Furthermore, men, older age group >30 years, HIV positive status, anemia (hemoglobin ≤10 mg/dl), and history of weight loss at presentation were significantly associated with increased relative hazards of death (Table 9:7).

In multivariable Cox proportional hazards model for anthropometric data, after adjusting for age, HIV, current smoking status, history of weight loss, and chest X-ray disease extent, the relative hazard for death when patients had reduced BMI <18.5 kg/m2 at presentation was 1.85 (95% CI, 1.25, 2.73; Table 9:8 and 9:9). Basing on our prior findings and the descriptive statistics presented in this paper that body composition differ

253

by gender (Mupere et al. 2010), we fitted separate nested Cox regression models with the same covariates according to gender (Table 9:9). Men with reduced BMI <18.5 kg/m2 at presentation had a significantly 1.70 (95% CI: 1.03, 2.81) times greater relative hazard of death than men with normal BMI ≥18.5 kg/m2. A 1.83 (95% CI: 0.96, 3.50) relative

hazard of death was found among women presenting with reduced BMI compared to

women presenting with normal BMI; however, this was not significant. The differences

in the relative hazard of death suggest interaction across the gender stratum. Older

patients >30 years of age among men presenting with reduced BMI also had significantly

greater relative hazard of death compared to younger men. In addition, HIV positive and patients with history of weight loss presenting with reduced BMI had significantly greater relative hard of death compared to HIV negative and patients without history of weight loss regardless of gender, respectively (Table 9:9).

In a multivariable model with reduced BMI as main predictor and after adjusting for age, sex, prior smoking status, history of weight loss, and extent of disease on chest x-ray,

patients with reduced BMI had a 1.75 (95% CI: 1.18, 2.60) times greater hazard of death

compared to individuals with normal BMI. Among HIV positive patients in stratified Cox

regression, individuals with reduced BMI had a significant 1.63 (95% CI: 1.09, 2.44)

times greater hazard of death compared to individuals with normal BMI whereas among

HIV negative patients, patients with reduced BMI had a 6.95 (95% CI: 0.78, 61.89) times

greater hazard of death compared to patients with normal BMI (Table 9:10 and 9:11).

This suggests interaction according to HIV status when patients present with/or without

reduced BMI. There were also potential interactions according to gender, age group, and

254

prior history of smoking. HIV positive men, HIV positive older patients, and HIV positive patients with prior history of smoking had significant hazards of death compared to HIV positive women, HIV positive young patients, and HIV positive patients with no prior history of smoking (Table 9:11). The converse was true that HIV negative men,

HIV negative older patients, and HIV negative patients with prior history of smoking had similar hazards of death compared to HIV negative women, HIV negative young patients, and HIV negative patients with no prior history of smoking. There were no significant interactions between the reduced BMI variable for wasting and all the variables adjusted for in the model (Appendix, Table 13:4). However, interactions with male sex, positive

HIV status, and older age group were found to be important in the reduced BMI multivariable model (Appendix, Table 13:5).

In a multivariable model with reduced BMI as main predictor and after adjusting for HIV positive status, male sex, prior smoking status, history of weight loss, and extent of disease on chest x-ray, patients with reduced BMI had a 1.78 (95% CI: 1.20, 2.64) times greater hazard of death compared to patients with normal BMI. Among young patients in stratified Cox regression, individuals with reduced BMI had a significant 2.85 (95% CI:

1.41, 5.77) times greater hazard of death compared to individuals with normal BMI whereas among old patients, individuals with reduced BMI had a 1.39 (95% CI: 0.86,

2.24) times greater hazard of death compared to patients with normal BMI (Appendix,

Tables 13:6, 13:7, 13:8, and 13:9). This also suggests interaction according to age group stratum of young (≤30 years) and old (>30 years) when patients present with/or without reduced BMI.

255

In the multivariable Cox proportional hazards model using reduced FFMI as main predictor after adjusting for age, HIV, and anemia (hemoglobin ≤10 mg/dl) the relative hazard for death when patients had reduced FFMI (<16.7 kg/m2 for men and <14.6 kg/m2 for women) at presentation was 1.88 (95% CI, 0.96, 3.65; Table 9:12 and 9:13). When sex, HIV, age, and hemoglobin were included in this model, there was a statistically significant interaction between patients presenting with reduced FFMI and gender

[hazard ratio (HR), 0.122 (95% CI, 0.030, 0.56)]. Furthermore, basing on our prior findings that body composition differ by gender (Mupere et al. 2010), we fitted separate stratified Cox regression models with the same covariates according to gender (Table

9:13). Women with reduced FFMI <14.6 kg/m2 at presentation had nearly 7 fold greater relative hazard of death than women with normal FFMI ≥14.6 kg/m2. While men presenting with reduced FFMI <16.7 kg/m2 had a protective hazard of death of 0.80 (95%

CI, 0.35, 1.84) compared to men presenting with normal FFMI ≥16.7 kg/m2, although this was not significant. Older patients >30 years of age among women presenting with reduced FFMI also had significantly greater relative hazard of death compared to young women. In addition, HIV positive patients presenting with reduced FFMI had significantly greater relative hazard of death compared to HIV negative patients regardless of gender (Table 9:13).

Interactions of reduced FFMI variable with HIV, age, and hemoglobin in the model were not significant; however, they were found to be important in the model (Appendix,

Tables 13:12 and 13:13). In a multivariable model with reduced FFMI as main predictor and after adjusting for age, sex, and hemoglobin, individuals with reduced BMI had a

256

1.50 (95% CI: 0.73, 3.08) times greater hazard of death compared to individuals with

normal FFMI. Among HIV positive patients in stratified Cox regression, individuals with

reduced FFMI had a non-significant 1.50 (95% CI: 0.72, 3.12) times greater hazard of

death compared to individuals with normal FFMI whereas among HIV negative patients,

the hazard of death for presenting with reduced FFMI could not be estimated because of

very events (Appendix, Tables 13:14 and 13:15). This suggests that HIV is not a

confounder when patients present with/or without reduced BMI.

In a multivariable model with reduced FFMI as main predictor and after adjusting for

HIV positive status, male sex, and hemoglobin, patients with reduced BMI had a non-

significant 1.55 (95% CI: 0.76, 3.15) times greater hazard of death compared to patients

with normal FFMI. Among young patients in stratified Cox regression, individuals with

reduced FFMI had a non-significant 1.89 (95% CI: 0.55, 6.52) times greater hazard of

death compared to individuals with normal FFMI. Similarly, among old patients,

individuals with reduced FFMI had a non-significant 1.52 (95% CI: 0.63, 3.69) times greater hazard of death compared to patients with normal FFMI (Appendix, Tables 13:13

– 13:16). This also suggests that age is not a confounder when patients present with/or without reduced FFMI.

There was no significant relative hazard of death between fat mass body wasting and any of the main effect variables in both univariate, multivariable, and stratified Cox models.

257

Discussion

To our knowledge, this is the first study to report the effect of body composition on

survival among tuberculosis patients and the involved gender differences. In this

retrospective cohort study of adult patients with pulmonary tuberculosis in urban Uganda,

body wasting as measured by reduced fat-free mass and body mass indexes was

associated with reduced survival, but the effect varied according to gender of the patient.

Among men with reduced BMI <18.5 kg/m2 at diagnosis, the effect of body mass wasting

was substantial whereas in women with reduced BMI, the effect was minimal. Among

women with reduced fat-free mass index (FFMI <14.6 kg/m2) at diagnosis, the effect of

fat-free mass wasting was dramatic with a 7 fold risk of death whereas among men presenting with reduced FFMI, the effect on observed survival was protective. The current results indicate that reduced body mass is a predictor of survival among men whereas reduced fat-free mass is a predictor of survival among women and the effect of

wasting is greatest early in the first year following diagnosis of tuberculosis. Fat mass

wasting appears not to be predictor of survival regardless of gender.

This study found gender differences in the measures of body wasting and malnutrition

and the corresponding effect on observed survival. Compared to FFMI, BMI was found

to be a better measure of body wasting associated with the observed survival in men

whereas FFMI was better measure in women. This appears to suggest that BMI may

underestimate mortality in women whereas FFMI may underestimate mortality in men. A

possible explanation stems from the fact that BMI is composed of both fat and fat-free

258

mass each of which influence BMI in the same direction concerning mortality as

previously hypothesized (Allison et al. 1997). However, fat and fat-free mass may influence mortality in opposite directions; that is, fat-free mass may be protective and fat mass deleterious. Though BMI is monotonically related to adiposity (Garrow and

Webster 1985), it also correlates positively with the amount of fat-free mass an individual has. In our study, BMI had a significant positive correlation of 0.85 in men and 0.58 in women. Thus, the inverse gender differences in fat and fat-free mass explains the differences in correlation. The high hazard of death reported in this study among men and among women may be an indicator of excessively low amount of fat mass in men and low amount of fat-free mass in women regardless of BMI. However, body fat mass wasting was not associated with risk of death regardless of among women and men.

The findings of the present appear to suggest that the effects of body wasting and malnutrition on tuberculosis are not apparent when there is no comorbidity, yet tuberculosis is associated with malnutrition (Macallan 1999; Schwenk A and Macallan

D.C 2000). In the face of comorbidities such as HIV and anemia, the effects of body wasting and malnutrition become detectable because the force of mortality from comorbidity overwhelms the existing malnutrition. Results of the present study and similar to previous studies in sub-Saharan Africa (Shah et al. 2001; Zachariah et al.

2002), show that more than 30% of tuberculosis patients present with body wasting and malnutrition regardless of gender or HIV status, and whether BMI or BIA parameters were used in assessing nutritional status. Furthermore in the present study, there were no differences in body wasting between HIV positive and HIV negative patients at the time

259

of diagnosis suggesting that tuberculosis may be the driving factor in inducing the

wasting process and the role of HIV is minimal. This finding is consistent with previous

studies (Mupere et al. 2010; Paton and Ng 2006). However overtime in the present study, results revealed heterogeneity in the effects of HIV on survival in the presence of body

wasting as measured by reduced BMI. Patients that presented with reduced BMI (<18.5

kg/m2) and were HIV positive had a significant 1.63 (95% CI: 1.09, 2.44) hazard of death

compared to patients that presented with normal BMI and were HIV positive. Yet there

was minimal effect on observed survival among patients that presented with reduced BMI

and were HIV negative. Malnutrition in itself is a cause of immunodepression

(Hernandez-Pando, Orozco, and Aguilar 2009); thus, tuberculosis might worsen the

course of HIV-associated immunodepression. These interrelated effects possibly explain

the facts that both tuberculosis and malnutrition are associated with reduced survival

among HIV-infected patients (Nunn et al. 1992; Suttmann et al. 1995).

In this study, more than 75% of deaths occurred early in the first year of follow-up. There are several potential reasons for the early deaths of tuberculosis patients in the present study. First, late presentation with severe and extensive tuberculosis; more than 75% of the participants had moderate/or far advanced disease on chest x-ray. Second, most

(>90%) deaths were HIV-related; yet HIV-tuberculosis co-infection is associated with extra burden of nutritional alterations that may lead to poor outcomes. The extra burden on nutritional status may include increased energy expenditure, nutrient malabsorption, reduced intake, micronutrient malnutrition, and increased production of inflammatory cytokines with lipolytic and proteolytic activity (van Lettow, Fawzi, and Semba 2003;

260

Niyongabo et al. 1994; Melchior et al. 1993). Third, about 25% of the study population

presented with anemia (hemoglobin ≤10 mg/dl); yet anemia is a life-threatening HIV-

related complication that may be associated with poor outcome. Finally, wasting status at

diagnosis; we have shown that body wasting regardless of whether assessed by BMI or

fat-free mass index was associated with an increased hazard of death which could be early or late.

Findings in the present study show gender differences in survival and the survival was modified by HIV infection. HIV positive men had poor survival compared to HIV positive women; however, HIV negative men and HIV negative women had similar survival. Our findings suggest that in the face of co-infection, survival is poor among

HIV positive men while in the absence of HIV, survival is similar by gender. Our findings are similar to previous studies of mortality in HIV-infected patients in Africa in which the rate of mortality was higher in men than in women (Lucas et al. 1993; Sani et al. 2006). Our study however, had a large sample size and a heterogeneous study population that included ambulatory and inpatients compared to previous studies (Sani et al. 2006; Lucas et al. 1993) enrolled only in-patients with probable advanced disease. The poor survival among HIV positive men could be due to delay in presentation for care among men. Prior studies have explained the late presentation among men to be due to a false sense of security concerning relative risk HIV infection (Luginaah, Yiridoe, and

Taabazuing 2005). Further studies are needed to evaluate the socioeconomic and biologic factors that predispose HIV positive men to poor survival.

261

The caveat to the interpretation of findings in this study include 1) the method (BIA) used in measurement of body composition is not of reference standard like the dual-energy x- ray absorptiometry, 2) the BIA prediction method used has not yet been validated in the local population. As a result, findings of body composition may be biased because of variations in hydration across ethnic groups (Kyle et al. 2004). However, the equations that were used in this study were previously cross validated in individuals of different race (white, black, and Hispanic) among men and women, who were both healthy controls and HIV-infected patients (Kotler et al. 1996). Moreover, the equations have been used widely in other studies from Africa with meaningful findings (Shah et al. 2001;

Van Lettow et al. 2004; Villamor et al. 2006; Mupere et al. 2010). We also took care to take measurements at rest, with proper placement of leads, in participants who had not exercised or taken alcohol, in participants with voided bladder and ambient temperature.

However, measurements were in patients with underlying illness that may cause shifts in body water compartments, thereby affecting measurements of fat mass. Our findings also limited by the lack of dietary intake assessment to give further insight in the interpretation of gender differences in body composition and observed survival.

Despite the limitations of the present study, the strengths of this study are that a large number of patients were studied, long duration of follow-up period to observe survival, and assessing of nutritional status using both BMI and the precise measures of fat and fat- free mass.

262

The findings in this study indicate that body wasting exerts greatest effect on observed survival among tuberculosis patients with body wasting co-infected with HIV, and that

BMI is a better predictor of death among men whereas FFMI is a better predictor of death among women. These observations provide evidence for the need to provide nutritional interventions among HIV co-infected patients with wasting and evaluation of nutritional status should be involve BMI, fat and fat-free mass.

Acknowledgements

We thank all study staff members of the Case Western Reserve University and Makerere

University research collaboration at the Tuberculosis Research Unit in the United States and in Uganda for their assistance; the faculty of staff at Case Western Reserve

University Department of Epidemiology and Biostatistics for the guidance in analyzing the project; and the Fogarty International Center, for the continued support.

This study was supported in part by the AIDS International Training Research Program,

Fogarty International Center, Grant No. TW000011, based at Case Western Reserve

University, Department of Epidemiology and Biostatistics, and Tuberculosis Research

Unit (grant N01-AI95383 and HHSN266200700022C/ N01-AI70022 from the NIAID).

This work was part of Ezekiel Mupere’s PhD thesis at Case Western Reserve University.

263

Table 9:2 Mean body mass, fat and fat-free mass indexes; spearman correlations

between body mass index and fat or fat-free mass indexes among adult women and men in urban Uganda

Mean (SD)a Correlations with BMIb

Characteristic Women Men Women Men

(n=147) (n=164) (n=147) (n=164)

1BMI, kg/m2 20.0 (3.3) 18.6 (2.1) 1 1

FFMI, kg/m2 15.8 (1.1) 16.6 (1.4) 0.58 0.85

FMI, kg/m2 4.4 (2.8) 2.0 (0.8) 0.92 0.68

BMI = body mass index, FFMI = fat-free mass index, FMI = fat mass index. 1Three hundred fifty two women and 401 men were involved in estimation of mean BMI; aall p- values were <0.001 for mean differences between women and men; ball correlations between FFMI or FMI and BMI among women and men were significant with p-values <0.001.

264

Table 9:3 Concordance between low body mass index and low fat-free or fat mass indexes corresponding to body mass index in assessing wasting among adults in urban Uganda

Women

Characteristic Kappa, к Normal BMI aReduced BMI (95% CI)

FFMI index kg/m2b

Reduced, n (%) 5/100 (5) 16/47 (34) 0.34 (0.18, 0.50)

Normal, n (%) 95/100 (95) 31/47 (66)

FMI index kg/m2b

Reduced, n (%) 0/47 (0) 47/47 (100) 0.61 (0.49, 0.73)

Normal, n (%) 71/100 (71) 29/100 (29)

Men

FFMI kg/m2b

Reduced, n (%) 15/76 (20) 61/76 (80) 0.57 (0.45, 0.69)

Normal, n (%) 71/92 (77) 21/92 (23)

FMI kg/m2b

265

Reduced, n (%) 27/76 (36) 49/76 (64) 0.52 (0.39, 65)

Normal, n (%) 80/92 (87) 12/92 (13)

FFMI = fat-free mass index, FMI = fat mass index, BMI = body mass index. aReduced BMI = <18.5 kg/m2. bLow FFMI for women <14.6 kg/m2 and for men <16.7 kg/m2; low FMI for women <3.9 and for men <1.8 kg/m2.

266

Table 9:4 Baseline characteristics of pulmonary tuberculosis patients with normal versus malnutrition

Normal Low Normal Low

Characteristic BMI BMI FFMI FFMI

(n=437) (n=310) (n=208) (n=103)

n (%) n (%) n (%) n (%)

Sex

Females 236 (54) 116 (37)a 126 (61) 21 (20) a

Males 201 (46) 194 (63) 82 (39) 82 (80)

Age (years)

≤30 257 (59) 177 (57) 135 (65) 59 (57)

>30 1880 (41) 133 (43) 73 (35) 44 (43)

HIV status

Negative 202 (46) 143 (46) 116 (56) 56 (54)

Positive 234 (54) 167 (54) 91 (44) 47 (46)

Hemoglobin (g/dl)1

>10 216 (81) 126 (64)a 151 (76) 73 (74)

267

≤10 51 (19) 70 (36) 48 (24) 26 (26)

Fat mass index (kg/m2)2

Normal 149 (79) 27 (22)a 124 (61) 52 (51)

Low 39 (21) 95 (78) 79 (39) 49 (49)

Chest x-ray disease extent3 75 (17) 36 (12)b 39 (19) Normal/mild 10 (10)b

Moderate/far advanced 362 (83) 274 (88) 169 (81) 93 (90)

Smoker4

No 374 (86) 219 (71)a 176 (85) 62 (60)a

Yes 61 (14) 90 (29) 31 (15) 41 (40)

Takes alcohol5

No 266 (61) 201 (65) 151 (73) 64 (62)b

Yes 170 (39) 109 (35) 56 (27) 39 (38)

History weight loss6

No 111 (26) 54 (17)b 38 (18) 13 (13)

Yes 324 (74) 255 (83) 168 (82) 90 (87) ap-value <0.001, bp-value <0.05, FFMI = fat-free mass index, BMI = body mass index. 1284 missed hemoglobin measurement due to lack of blood; 2FFMI and FFMI was not measured; 3eight missed extent variable; 4three missed history of ever smoked; 5one

268

missed history of alcohol intake; 6four missed history of weight loss; FFMI low <16.7 kg/m2 for men and <14.6 kg/m2 for women, normal ≥ 16.7kg/m2 for men, ≥14.6 kg/m2 for women; low BMI <18.5 kg/m2 for men and women.

269

Table 9:5 Stratified Analysis of mortality among tuberculosis with normal (≥18.5) and low (<18.5) body mass index (BMI kg/m2) according to key variables

[deaths/number at risk (%)]

Normal BMI Low BMI

Category (n=437) (n=310) RR of death (95% CI)

n (%) n (%)

Deaths

No 390 (89) 252 (81) 1.74 (1.22, 2.48)

Yes 47 (11) 58 (19)

Sex

Female 21/236 (9) 17/116 (15) 1.65 (0.90, 3.00)

Male 26/201 (13) 41/194 (21) 1.63 (1.04, 2.56)

Age group

≤30 years 13/257 (5) 21/177 (12) 2.35 (1.21, 4.56)

>30 years 34/180 (19) 37/133 (28) 1.47 (0.98, 2.22)

HIV-serostatus1

Negative 1/202 (0.5) 5/143 (3.5) 7.06 (0.83, 59.81)

270

Positive 46/234 (20) 53/167 (32) 1.61 (1.5, 2.27)

Hemoglobin2

>10 mg/dl 13/216 (6) 17/126 (13) 2.24 (1.13, 4.46)

≤10 mg/dl 10/51 (20) 16/70 (23) 1.17 (0.58, 2.35)

Smoker3

No 41/374 (11) 43/219 (20) 1.79 (1.21, 2.66)

Yes 6/61 (10) 14/90 (16) 1.58 (0.64, 3.89)

Takes alcohol4

No 25/266 (9) 39/201 (19) 2.06 (1.29, 3.30)

Yes 22/170 (13) 19/109 (17) 1.35 (0.77, 2.37)

Weight loss5

No 5/111 (5) 6/54 (11) 2.47 (0.79, 7.72)

Yes 42/324 (13) 52/255 (20) 1.57 (1.08, 2.28)

Chest x-ray extent6

Normal/minimal 9/75 (12) 10/36 (28) 2.31 (1.03, 5.19)

Moderate/far advanced 38/357 (11) 48/271 (18) 1.66 (1.12, 2.47)

1One missed HIV status; 2284 missed hemoglobin measurement due to lack of blood; 6eight missed extent variable; 3three missed history of ever smoked; 4one missed history of alcohol intake; 5four missed history of weight loss; low BMI <18.5 kg/m2.

271

Table 9:6 Stratified Analysis of mortality among tuberculosis with normal and low

fat-free mass index (FFMI kg/m2) according to key variables [deaths/number at risk

(%)]

Normal FFMI Low FFMI

Category (n=208) (n=103) RR of death (95% CI)

Deaths

No 194 (93) 87 (84) 2.31 (1.17, 4.54)

Yes 14 (7) 16 (16)

Sex

Female 6/126 (5) 6/21 (29) 6.00 (2.14, 16.86)

Male 8/82 (10) 10/82 (12) 1.25 (0.52, 3.01)

Age group

≤30 years 5/135 (4) 5/59 (8) 2.29 (0.69, 7.61)

>30 years 9/73 (12) 11/44 (25) 2.03 (0.91, 4.50)

HIV-serostatus1

Negative 0/116 (0) 1/56 (1.8) -

Positive 14/91 (15) 15/47 (32) 2.07 (1.10, 3.92)

272

Hemoglobin2

>10 mg/dl 8/160 (5) 7/77 (9) 1.82 (0.68, 4.83)

≤10 mg/dl 6/48 (13) 9/26 (35) 2.79 (1.11, 6.92)

Smoker3

No 12/176 (7) 11/62 (18) 2.60 (1.21, 5.59)

Yes 2/31 (6) 5/41 (12) 1.89 (0.39, 9.10)

Takes alcohol4

No 10/151 (7) 9/64 (14) 2.12 (0.91, 4.98)

Yes 4/56 (7) 7/39 (18) 2.51 (0.79, 8.00)

Weight loss5

No 2/38 (5) 0/13 (0) -

Yes 12/168 (7) 16/90 (18) 2.49 (1.23, 5.03)

Chest x-ray extent6

Normal/minimal 2/39 (5) 5/10 (50) 9.75 (2.21, 43.06)

Moderate/far advanced 12/164 (7) 11/93 (12) 1.62 (0.74, 3.52)

1One missed HIV status; 215 missed hemoglobin measurement due to lack of blood; 6eight missed extent variable; 3one missed history of ever smoked; 4one missed history of alcohol intake; 5four missed history of weight loss; FFMI low (<16.7 kg/m2 for men and <14.6 kg/m2 for women, normal (≥ 16.7kg/m2 for men, ≥14.6 kg/m2 for women).

273

Table 9:7 Univariate Analysis of factors associated with survival

Characteristic Deaths/N (%) Relative hazard (95% CI)

Fat-free mass index

(kg/m2)a

Normal 14/208 (7) 1

Low 16/103 (16) 2.34 (1.14, 4.80)

Fat mass index (kg/m2)a

Normal 16/176 (9) 1

Low 14/135 (10) 1.23 (0.60, 2.52)

Body mass index (kg/m2)

Normal 47/437 (11) 1

Low 58/310 (19) 1.80 (1.23, 2.64)

Sex

Female 38/352 (11) 1

Male 67/395 (17) 1.66 (1.12, 2.48)

Age (years)

≤30 34/434 (8) 1

274

>30 71/313 (23) 3.15 (2.09, 4.74)

HIV-serostatus1

Negative 6/345 (2) 1

Positive 99/401 (25) 16.24 (7.13, 37.02)

Hemoglobin (g/dl)2

>10 30/342 (9) 1

≤10 26/121 (21) 1.63 (1.00, 2.69)

Smoker3

No 84/593 (14) 1

Yes 20/151 (13) 0.98 (0.60, 1.60)

Takes alcohol4

No 64/467 (14) 1

Yes 41/279 (15) 1.10 (0.75, 1.63)

Chest x-ray extent6

Normal/minimal 19/111 (17) 1

Moderate/far advanced 86/628 (14) 0.73 (0.44, 1.19)

Weight loss5

275

No 11/165 (7) 1

Yes 94/579 (16) 2.34 (1.44, 4.80)

FFMI = fat-free mass index, BMI = body mass index, 95% CI = confidence interval. 1One missed HIV status; 2284 missed hemoglobin measurement due to lack of blood; 2fat and fat-free mass was not measured; 6eight missed extent variable; 3three missed history of ever smoked; 4one missed history of alcohol intake; 5four missed history of weight loss; a311 had FMI and FFMI, respectively); FFMI low <16.7 kg/m2 for men and <14.6 kg/m2 for women, normal ≥ 16.7kg/m2 for men, ≥14.6 kg/m2 for women; low BMI <18.5 kg/m2 for men and women.

276

Table 9:8 Relative hazards [HR, 95% confidence intervals (CIs)] for death among tuberculosis patients with normal compared with patients having low body mass index (BMI)

Overall model Characteristics Deaths/N (%)

HR (95% CI)

BMI (kg/m2)

Normal (≥18.5) 47/437 (11) 1

Low (<18.5) 58/310 (19) 1.85 (1.25, 2.73)

Age (years)

≤30 34/434 (8) 1

>30 71/313 (23) 2.31 (1.53, 3.49)

HIV status1

Negative 6/345 (2) 1

Positive 99/401 (25) 15.86 (6.92, 36.32)

Smoker2

No 84/593 (14) 1

277

Yes 20/151 (13) 0.69 (0.42, 1.14)

Weight loss3

No 11/165 (7) 1

Yes 94/579 (16) 3.40 (1.81, 6.37)

Chest x-ray extent4

Normal/minimal 19/111 (17) 1

Moderate/far advanced 86/628 (14) 0.80 (0.49, 1.32)

1One missed HIV status; 4eight missed extent variable; 2three missed history of ever smoked; 3four missed history of weight loss; low BMI <18.5 kg/m2.

278

Table 9:9 Relative hazards [HR, 95% confidence intervals (CIs)] for death among tuberculosis patients with normal compared with patients having low body mass index (BMI) stratified according to sex status

Stratified models

Deaths/N Characteristics Women (n=352) Men (n=395) (%)

HR (95% CI) HR (95% CI)

BMI (kg/m2)

Normal (≥18.5) 47/437 (11) 1 1

Low (<18.5) 58/310 (19) 1.83 (0.96, 3.50) 1.70 (1.03, 2.81)

Age (years)

≤30 34/434 (8) 1 1

>30 71/313 (23) 1.77 (0.93, 3.37) 2.57 (1.46, 4.54)

HIV status1

Negative 6/345 (2) 1 1

Positive 99/401 (25) 12.61 (3.84, 41.37) 18.48 (5.76, 59.26)

Smoker2

No 84/593 (14) 1 1

279

Yes 20/151 (13) 0.38 (0.05, 2.78) 0.71 (0.37, 1.37)

Weight loss3

No 11/165 (7) 1 1

Yes 94/579 (16) 2.60 (1.00, 6.76) 4.27 (1.84, 9.95)

Chest x-ray extent4

Normal/minimal 19/111 (17) 1 1

Moderate/far advanced 86/628 (14) 0.91 (0.42, 1.99) 0.61 (0.35, 1.04)

1One missed HIV status; 4eight missed extent variable; 2three missed history of ever smoked; 3four missed history of weight loss; low BMI <18.5 kg/m2.

280

Table 9:10 Relative hazards [HR, 95% confidence intervals (CIs)] for death among tuberculosis patients with normal compared with patients having low body mass index (BMI)

Overall model Characteristics Deaths/N (%)

HR (95% CI)

BMI (kg/m2)

Normal (≥18.5) 47/437 (11) 1

Low (<18.5) 58/310 (19) 1.75 (1.18, 2.60)

Age (years)

≤30 34/434 (8) 1

>30 71/313 (23) 3.10 (2.05, 4.69)

Gender

Female 38/352 (11) 1

Male 67/395 (17) 1.63 (1.07, 2.50)

Smoker2

No 84/593 (14) 1

281

Yes 20/151 (13) 0.56 (0.33, 0.93)

Weight loss3

No 11/165 (7) 1

Yes 94/579 (16) 2.70 (1.44, 5.07)

Chest x-ray extent4

Normal/minimal 19/111 (17) 1

Moderate/far advanced 86/628 (14) 0.66 (0.40, 1.09)

1One missed HIV status; 4eight missed extent variable; 2three missed history of ever smoked; 3four missed history of weight loss; low BMI <18.5 kg/m2.

282

Table 9:11 Relative hazards [HR, 95% confidence intervals (CIs)] for death among tuberculosis patients with normal compared with patients having low body mass index (BMI) stratified according to HIV status

Stratified models

HIV negative HIV positive Characteristics Deaths/N (%) (n=402) (n=345)

HR (95% CI) HR (95% CI)

BMI (kg/m2)

Normal (≥18.5) 47/437 (11) 1 1

Low (<18.5) 58/310 (19) 6.95 (0.78, 61.89) 1.63 (1.09, 2.44)

Age (years)

≤30 34/434 (8) 1 1

>30 71/313 (23) 4.89 (0.84, 28.39) 2.10 (1.37, 3.22)

Gender

Female 38/352 (11) 1 1

Male 67/395 (17) 0.57 (0.10, 3.31) 1.62 (1.05, 2.52)

Smoker2

No 84/593 (14) 1 1

283

Yes 20/151 (13) 1.20 (0.18, 8.17) 0.56 (0.33, 0.96)

Weight loss3

No 11/165 (7) 1 1

Yes 94/579 (16) - 3.46 (1.84, 6.49)

Chest x-ray extent4

Normal/minimal 19/111 (17) 1 1

Moderate/far advanced 86/628 (14) - 0.73 (0.44, 1.21)

1One missed HIV status; 4eight missed extent variable; 2three missed history of ever smoked; 3four missed history of weight loss; low BMI <18.5 kg/m2.

284

Table 9:12 Relative hazards [HR, 95% confidence intervals (CIs)] for death among tuberculosis patients with normal compared with low fat-free mass index (FFMI)

Overall model Characteristics Deaths/N (%)

HR (95% CI)

FFMI in kg/m2

Normal (≥18.5) 14/208 (7) 1

Low (<18.5) 16/103 (16) 1.88 (0.96, 3.65)

Age (years)

≤30 11/212 (5) 1

>30 24/128 (19) 2.57 (1.26, 5.28)

HIV-serostatus1

Negative 1/186 (0.5) 1

Positive 34/153 (22) 34.43 (4.66, 254.32)

Hemoglobin (mg/dl)2

>10 18/242 (7) 1

285

≤10 16/83 (47) 1.68 (0.86, 3.28)

1One missed HIV status; 215 missed hemoglobin measurement due to lack of blood; FFMI low (<16.7 kg/m2 for men and <14.6 kg/m2 for women, normal (≥ 16.7kg/m2 for men, ≥14.6 kg/m2 for women).

286

Table 9:13 Relative hazards [HR, 95% confidence intervals (CIs)] for death among

tuberculosis patients with normal compared with low fat-free mass index (FFMI) stratified according to sex status

Stratified models

Characteristics Deaths/N (%) Women (n=165) Men (n=175)

HR (95% CI) HR (95% CI)

FFMI in kg/m2

Normal (≥18.5) 14/208 (7) 1 1

Low (<18.5) 16/103 (16) 6.83 (2.14, 21.74) 0.80 (0.35, 1.84)

Age (years)

≤30 11/212 (5) 1 1

>30 24/128 (19) 3.54 (1.09, 11.47) 2.22 (0.87, 5.66)

HIV-serostatus1

Negative 1/186 (0.5) 1 1

Positive 34/153 (22) - 21.44 (2.84, 161.99)

Hemoglobin (mg/dl)2

>10 18/242 (7) 1 1

287

≤10 16/83 (47) 0.73 (0.23, 2.32) 2.88 (1.26, 6.57)

1One missed HIV status; 215 missed hemoglobin measurement due to lack of blood; FFMI low (<16.7 kg/m2 for men and <14.6 kg/m2 for women, normal (≥ 16.7kg/m2 for men, ≥14.6 kg/m2 for women).

288

Figure 9:1 Survival distribution among adult men presenting with wasted body mass (BMI <18.5 kg/m2) compared to men with normal body mass in urban

Kampala, Uganda

289

Figure 9:2 Survival distribution among adult women with baseline fat-free mass wasting (FFMI <16.7 kg/m2 for men, <14.6 kg/m2 for women) compared to women with normal fat-free mass in urban Uganda

290

Figure 9:3 Survival distribution Men compared with Women among HIV positive tuberculosis patients in urban Uganda

Data truncated at 1462 days of follow-up

291

CHAPTER 10

LONGITUDINAL CHANGES IN BODY COMPOSITION AMONG HIV

POSITIVE AND HIV NEGATIVE ADULT PATIENTS WITH PULMONARY

TUBERCULOSIS IN URBAN KAMPALA, UGANDA

292

Abstract

Background Follow-up studies are limited to understand nutritional changes during and after tuberculosis treatment. The effect of body wasting at the time of diagnosis and HIV infection on rate of change for fat-free mass, fat mass, and body mass index (BMI) during and after treatment were evaluated among tuberculosis patients in urban Kampala,

Uganda.

Methods In a retrospective cohort study of 717 adult patients, BMI was assessed at baseline, 2, 3, 5, 6, 12, and 24 months whereas fat-free mass index (FFMI) and fat mass index (FMI) were evaluated at baseline, 3, 12, and 24 months. Longitudinal mixed effects two piecewise models with knots at month 3 and month 12 were fit to the data.

Results There were no differences in body wasting as assessed by reduced FFMI, FMI, and BMI between HIV positive and HIV negative patients at presentation. In stratified mixed effects two spline models during the first three months of treatment, the gain in

FFMI among patients that presented with reduced fat-free mass at diagnosis was dramatic in men with rate of 4.55 kg/m2 (95% confidence interval (CI): 1.26, 7.83); however, the

gain was minimal among women who presented with reduced FFMI with rate of 2.07

kg/m2 (95% CI: -0.74, 4.88). In stratified models for FMI as dependent variable, women

with reduced fat mass at presentation had a substantial gain in FMI at rate of 3.55 kg/m2

(95% CI: 0.40, 6.70) whereas men had a rate of 3.16 kg/m2 (0.80, 5.52). In stratified

models with BMI as dependent variable, men with reduced BMI at presentation gained

BMI at a rate of 6.45 kg/m2 (95% CI: 3.02, 9.87) whereas women at a rate of 3.30 kg/m2

(95% CI: -0.11, 6.72). There were minimal changes in FFMI, FMI, and BMI during the

293

first three months of treatment in stratified models according to HIV status. Furthermore, there were minimal changes in FFMI, FMI, and BMI after month 3 and during the one year follow-up after month 12.

Conclusion Gender but not HIV was associated with longitudinal body composition changes during the initial phase of treatment among tuberculosis patients that presented with body wasting. Further evaluation is needed to under the impact of providing nutritional interventions as adjuvant treatment on body composition among tuberculosis patients in sub-Saharan.

294

Background

The global burden of tuberculosis remains enormous because there are insufficient

tuberculosis control programs and higher rates of tuberculosis and human

immunodeficiency virus (HIV) co-infection. The incidence remains high, with 9.2 million new cases reported to have occurred in 2006 only (Vitoria et al. 2009). The majority of these cases occurred in Asia, but the highest population rate has been observed in Africa where the prevalence of HIV infection is at its highest.

Body wasting is regarded as a cardinal feature of tuberculosis; however, the pathophysiology of wasting remains poorly understood (Schwenk A and Macallan D.C

2000). A significant proportion of African tuberculosis patients have a marked degree of wasting by the time they present for registration and treatment (Kennedy et al. 1996;

Harries et al. 1988; Zachariah et al. 2002). Wasting is associated with impaired physical function (Harries et al. 1988), longer hospitalization days and increased mortality in patients with tuberculosis (Rao et al. 1998; Zachariah et al. 2002; Mitnick et al. 2003).

Thus, early diagnosis of body wasting in tuberculosis is essential if timely interventions are to be instituted.

Wasting associated with tuberculosis is likely caused by a combination of decreased appetite and altered metabolism resulting from the inflammatory and immune responses

295

(Paton et al. 1999; Paton et al. 2003; Macallan 1999). The decreased appetite may lead to

a decrease in energy intake, interacting with increased losses with a resultant body

wasting. Although antituberculosis treatment is highly successful (Chemotherapy and

management of tuberculosis in the United Kingdom: recommendations 1998. Joint

Tuberculosis Committee of the British Thoracic Society 1998), many patients remain

underweight after 6 months of treatment (Onwubalili 1988) suggesting that full recovery may take a longer than the treatment itself.

It has been reported that in patients with both tuberculosis and HIV co-infection, the

wasting process is exacerbated (Macallan 1999; Lucas et al. 1994). In contrast, findings

from several cross-sectional studies appear to show no large differences in body

composition between HIV-infected adults with tuberculosis and HIV-negative adults with

tuberculosis at presentation (Shah et al. 2001; Niyongabo et al. 1999; Niyongabo et al.

1994; Mupere et al. 2010) suggesting that tuberculosis is the dominant factor inducing

wasting. This has been shown in several reports (Mupere et al. 2010; Paton and Ng

2006). However, the wasting process at the time of TB diagnosis according to previous

reports appear to differ by gender (Kennedy et al. 1996; Mupere et al. 2010).

Nutritional changes at the time of tuberculosis diagnosis are well documented using both

anthropometry and bioelectrical impedance analysis (BIA) in several reports (van Lettow,

Fawzi, and Semba 2003; Paton et al. 1999; Shah et al. 2001; Van Lettow et al. 2004;

Villamor et al. 2006; Mupere et al. 2010); however, follow-up studies are still limited to

296

understand nutritional changes during and after tuberculosis treatment. The limited follow-up studies (Kennedy et al. 1996; Ramakrishnan et al. 1961) found weight gain to be a poor indicator of clinical response in tuberculosis. Furthermore, weight or body mass index (BMI) provides limited information about nutritional alterations in patients with tuberculosis. In addition, BMI is insensitive to body fatness, particularly at low BMI, as well as with above-normal muscle development (Kyle, Genton, and Pichard 2002; Kyle,

Piccoli, and Pichard 2003). Measurement of body composition is necessary to obtain a true picture of the nutritional status in tuberculosis because body compartments differ in their contribution to weigh gain and its clinical benefit. Fat-free mass (FFM) and fat mass body composition measurements have been shown to permit a more precise evaluation of nutritional status (VanItallie et al. 1990; Kyle, Piccoli, and Pichard 2003). Fat-free mass

(FFM) is more closely correlated with quality of life and physical functioning than are fat mass (FM) and body weight (Wagner, Ferrando, and Rabkin 2000; Mostert et al. 2000).

BIA has been recommended as the preferred and precise method for clinical assessment of FFM and fat mass (Kyle et al. 2004; Kyle, Genton, and Pichard 2002). One follow-up study that evaluated FFM and FM had a small sample size, enrolled only HIV negative tuberculosis patients, and followed patients for only 6 months (Schwenk et al. 2004). In this large retrospective study, we evaluated the effect of baseline body wasting and HIV infection on the rate of change of FFM, FM, and BMI during and after treatment among tuberculosis patients in urban Uganda, Kampala.

297

Methodology

Study Design

A retrospective cohort study was conducted with the study population that consisted of

745 adult pulmonary tuberculosis patients having confirmed HIV status and defined baseline body wasting. The study used the completed five year Household Contact

(HHC) study, the completed phase II prednisolone double blind randomized placebo controlled clinical trial, and the ongoing Kawempe Community Health (KCH) study. Of the 745 patients, 314 were enrolled into the HHC, 341 into the KCH, and 90 into the placebo arm of the prednisolone clinical trial (Figure 10:1 and 10:2). The HHC and KCH studies were observational epidemiologic studies; organized and conducted by the

Makerere University and Case Western Reserve University tuberculosis research collaboration (Uganda-CWRU) that has been ongoing for the last 20 years in Uganda.

The HHC was the initial household contact study from 1995 to 1999 that described the epidemiology of tuberculosis in urban Kampala, Uganda (Guwatudde et al. 2003). The

KCH is the second phase of the HHC. The KCH phase started in 2002 and is still ongoing

(Stein et al. 2005). The KCH was developed specifically to focus on the determinants of host factors associated with primary infection, re-infection, reactivation, and progression of clinical disease and to identify and track individual strains of mycobacterial tuberculosis through Ugandan households and local community. The phase II clinical trial was conducted between 1995 to 2000 to determine whether immunoadjuvant prednisolone therapy in HIV-infected patients with TB who had CD4(+) T cell counts

>/=200 cells/ mu L was safe and effective at increasing CD4(+) T cell counts (Mayanja-

Kizza et al. 2005). Patients were eligible for this present study if they had baseline and

298

follow-up body measurements, had an HIV test, and were part of one prospective study

conducted by the Uganda-CWRU research collaboration.

The institutional review boards at Case Western Reserve University in the United States

and Joint Clinical Research Center in Uganda reviewed the protocol and final approval

was obtained from the Uganda National Council for Science and Technology. All

patients in the HHC and KCH had written informed consent to be enrolled in the study.

Biotech, Cambridge, MA). None of the HIV positive patients, neither those who were newly identified with HIV nor those with pre-existing HIV, were on antiretroviral therapy.

At enrollment, basic demographic information and a medical history were collected, and a standardized physical examination was conducted by a medical officer. Active tuberculosis was confirmed by sputum smear microscopy and culture. Patients with active tuberculosis were treated with standard four-drug chemotherapy for tuberculosis per guidelines of the Ugandan Ministry of Health. Adults with a previous history of treated pulmonary tuberculosis were excluded in the study. Of the 745 participants who were enrolled in the three studies, 28 were excluded due to lack of anthropometric follow-up data and being below 18 years of age, leaving 717 participants in total

299

available for analysis. The BIA data were collected during the KCH study only. Of the

341 participants who were enrolled in KCH, 63 were excluded due to lack of BIA follow-

up data and being below 18 years of age, leaving 278 participants in total available for

analysis (Figure 10:2). However, there were no differences in baseline age, gender,

weight, height, BMI, smoking status, hemoglobin, chest x-ray disease extent, and history

of weight loss between participants who were included and those who were excluded.

Measurements

Nutritional status was assessed using anthropometric measurements such as height and

weight and BIA Detroit, MI, RJL Systems. Body weight was determined to the nearest

0.1 kg using a SECA adult balance, and standing height was determined to the nearest 2 mm. Anthropometric measurements for the present retrospective study were performed during the HHC and KCH at scheduled visits including baseline, 2, 3, 5, 6, 12, and 24 months on follow-up. Body-mass index (BMI) was computed using the relationship of weight in kilograms divided by height in meters squared (kg/m2). All BIA measurements

were performed by one trained observer using the same equipment and recommended

standard conditions with regard to body position, previous exercise, dietary intake, skin

temperature, and voiding of the bladder were taken into consideration in taking BIA

measurements (Kyle et al. 2004). All BIA measurements during the KCH study were

performed on the day patients were confirmed to have tuberculosis disease and on

scheduled visits at 3, 12, and 24 months.

300

The BIA is a simple, easy, safe, non-invasive technique, that has been recommended for nutritional studies in the clinical setting (Kyle et al. 2004; Kyle, Piccoli, and Pichard

2003) and is a convenient method to determine the lean or fat-free mass and fat body compartments (Kyle, Piccoli, and Pichard 2003; Kyle et al. 2004). Single-frequency BIA was performed at 50 kHz and 800 mA with standard tetrapolar lead placement (Jackson et al. 1988). Before performing measurements on each participant, the BIA instrument was calibrated using the manufacturer’s recalibration device. The resistance and reactance were based on measures of a series circuit (Kotler et al. 1996). BIA measurements were performed in triplicate for each subject. Fat-free mass (FFM) was calculated from BIA measurements using equations that were previously cross-validated in a sample of patients (white, black and Hispanic) with and without HIV infection

(Kotler et al. 1996) and have been applied elsewhere in African studies (Villamor et al.

2006; Shah et al. 2001; Van Lettow et al. 2004). Fat mass (FM) was calculated as body weight minus FFM.

Definitions

We used BMI and height-normalized indices (adjusted for height2) of body composition that partition BMI into fat-free mass index (FFMI) and fat mass index (FMI) (Schutz,

Kyle, and Pichard 2002; VanItallie et al. 1990; Kyle, Piccoli, and Pichard 2003) to establish the body wasting status of participants. The FFMI and FMI have the advantages of compensating for differences in height and age (Kyle, Genton, and Pichard 2002).

Also, the use of the FFMI and FMI eliminates some of the differences between

301

population groups. We defined body wasting as patients having the low fat-free mass

index (FFMI) and the low body fat mass index (FMI) corresponding to WHO BMI

categories for malnutrition as previously reported (Table 10:1) (Kyle, Piccoli, and

Pichard 2003). The FFMI <16.7 (kg/m2) for men and <14.6 (kg/m2) for women and the

FMI <1.8 (kg/m2) for men and <3.9 (kg/m2) for women corresponds to a BMI of <18.5

kg/m2, the WHO cutoff for malnutrition (WHO Tech Rep 1995) among adults.

Table 10:1 Definitions of low and normal fat and fat-free mass index values for corresponding body mass index in adults

Characteristic Low Normal

Body mass index (BMI)a

Women and men in kg/m2 < 18.5 ≥ 18.5

Fat-free mass index (FFMI)b

Women in kg/m2 < 16.7 ≥ 16.7

Men in kg/m2 < 14.6 ≥ 14.6

Fat mass index (FMI)b

Women in kg/m2 < 1.8 ≥ 1.8

Men in kg/m2 < 3.9 ≥ 3.9

aWorld Health Organization categories, sex independent (WHO Tech Rep 1995). bKyle et al (Kyle, Piccoli, and Pichard 2003; Kyle et al. 2003).

302

Statistical analysis

Hypothesis

The hypothesis tested in analysis of this data was: the null hypothesis stated that there would be no difference in the rate of change for FFMI, FMI, and BMI between patients with baseline body wasting and patients with no baseline wasting; and between HIV positive and HIV negative patients during and after the course of tuberculosis treatment in urban Kampala, Uganda. The study hypothesis (alternative) was that patients with baseline wasting and HIV infection were associated with increased rates of FFMI, FMI, and BMI compared to patients with no baseline wasting and with HIV negative during and after treatment among tuberculosis patients in urban Kampala, Uganda.

Descriptive statistics

Baseline characteristics for participants with baseline wasting were compared with participants who had no baseline wasting using the χ2 test or Fisher’s exact test (where

expected counts were less than 5) for binary data and student’s t-test for continuous

variables or Wilcoxon-Mann Whitney test for variables not normally distributed.

Exploration of data

In order to gain insights for the covariate structure and model building, we explored data

by conducting correlation structure matrices for FFMI, FMI, and BMI; plots of individual

profile trajectories, boxplots, and mean profiles per exposure group of baseline wasting

303

and no wasting overtime. The raw correlation coefficients along and off the diagonals

were clearly unique for BMI suggesting a heterogeneous covariance structure. However,

correlation coefficients for both FFMI and FMI were nearly the same along the diagonals

and appear to decay with time suggesting stationary structure (Appendix, Table 14:1).

Plots of individual FFMI, FMI, and BMI profiles over time have different intercepts and

different positive slopes suggesting importance of random intercepts and slopes in the

models (Appendix, Figures 14:1 – 14:3). The boxplots and mean profiles for three indexes showed linear increasing trends over time suggesting inclusion of interaction terms between the indexes with time to be important (Appendix, Figures 14:4 – 14:15).

Missing observations

The 278 participants from the KCH study who were included in the BIA analysis contributed 1,112 observations, 205/1,112 (18%) of which were missing observations.

However, the 717 participants from all the three studies who were included in the anthropometric data analysis contributed 4,875 observations, 781/4,875 (16%) of which were missing observations (Figure 10:1). In order to understand the nature of missingness and whether the missing observations were nonignorable, we used the logistic regression approach for outcome-dependent missing (Ridout 1991). We performed separate generalizing estimating equations (GEE) models for FFM, FMI, and BMI using Proc

GENMOD with a logit link function to assess the effect of having wasting (low FFMI, low FMI, or BMI <18.5 kg/m2) versus no wasting (normal/high FFMI, normal/high FMI,

or BMI ≥18.5 kg/m2) at baseline on to the probability of missing FFM, FMI, or BMI

304

overtime adjusting factor for baseline sex, age, HIV status, history of weight loss, chest x-ray disease extent, and hemoglobin level. In both models of FFMI and FMI, for outcome dependent missing, the missing probability was not related to the baseline wasting and the lag measure (Appendix, Tables 14:2 and 14:3). Thus, we assumed the data missing not to be informative on the outcome variable and ignorable for FFMI, FMI and BMI.

Model building

The aim was to specify the appropriate mean and covariance structures of the data. The main outcome vectors of the participant’s FFMI, FMI, and BMI were modeled as linear combinations of baseline covariates and baseline body wasting status in separate models.

Changes in FFMI, FMI, or BMI over time were estimated using multilevel linear random effect/mixed effects models with random intercept and random slope, accounting for correlation of repeated measurements within each individual (Laird and Ware 1982;

Singer J.D and Willet J.B 2003; Bryk A.S and Ruandenbush S.W 2002). Multilevel modeling of change involves two levels: level one that describes how the outcome in each individual changes overtime (within-person change), and level two that describes how the within-person changes differ across persons (between-person differences in change) (Singer J.D and Willet J.B 2003). We used full maximum likelihood estimation method for parameter estimation and type 3 F-test for testing significance. We assumed a spatial exponential covariance structure. The spatial exponential covariance structure was plausible because the intervals between serial data points were different by design. The

305

exponential covariance structure is among the spatial correlation structures that takes the form: εij = Ui(tij) + еij. The Ui(tij) are assumed to have a normal distribution, with 0 mean,

(|tij – tik|) variance σ2 u, and correlation Corr { Ui(tij), Ui(tik)} = ρ . This correlation becomes

weaker as the separation increases. The еij are the usual sampling or measurement errors.

The LOCAL option was added to the REPEATED statement in SAS to decompose the

errors into Ui(tij) and еij.

The appropriateness of the covariance structure was assessed using Akaike Information

Criteria (AIC) and Bayesian Information Criteria (BIC) (Littell, Pendergast, and

Natarajan 2000). The quadratic and cubic terms were assessed using –2 log likelihood (-

2LL) and found to significantly improve the model (Appendix, Table 14:4). A model

with random intercepts and random slopes was assessed using -2LL and found to be

better than one without random intercepts or random slopes (Appendix, Table 14:4). We

also compared a piecewise model with two knots at month 3 and at month 12 with linear

and polynomial models. The knots in the piecewise were chosen a priori basing on the

scheduled data point measurements. The piecewise model was found to fit the data better

(Appendix, Table 14:5). Furthermore, a piecewise model with random intercepts and

random slopes was better than one without. We assessed for normality of the residuals at

modeling levels to identify the best function form of each variable included in the model.

We first fitted an unconditional mean model (outcome only, no predictors) to determine

whether the differences in average FFMI, FMI, and BMI across visits and persons were

306

non-zero and whether there was significant within- and between-person variability

(Singer J.D and Willet J.B 2003). After confirming non-zero variability in FFMI, FMI,

and BMI across visits and persons, and the existence of significant within-person and

between person variability, an unconditional growth model (with time as the only

predictor) was then fit to estimate the population average rate of FFMI, FMI, BMI

increase (slope) and initial status (intercept), and to determine if significant variability

existed in both statistics. The unconditional growth model showed significant variability

in both the initial status and the rate of change of FFMI, FMI, and BMI, suggesting the

need for predictor variables to explain the heterogeneity. The main predictor was baseline

body wasting using the following cutoffs alluded to above: low FFMI of 16.7 (kg/m2) for

men and 14.6 (kg/m2) for women, low FMI of 1.8 (kg/m2) for men and 3.9 (kg/m2) for

women (Kyle, Piccoli, and Pichard 2003; Schutz, Kyle, and Pichard 2002), and the

corresponding BMI cutoff of <18.5 kg/m2 (WHO Tech Rep 1995). The following were

other covariates used to assess associations with the dependent variables overtime:

number of visits in the study to adjust for missing data points, age categorized as young

(≤ 30 years) and old (> 30 years), sex, HIV sero-status, hemoglobin categorized as

anemic (≤ 10g/dl) and not anemic (>10), smoking status, history of weight loss, history of

taking alcohol now, and chest x-ray disease extent categorized as normal/minimal and moderate/far advanced. All covariates were time-invariant. The interval between study visits was adjusted to an annual scale to facilitate the interpretation of rates of FMI, FMI, and BMI decline per year.

307

We performed univariate analysis; variables with significant linear rates of FFMI, FMI, and BMI change and those with biological association with the three indexes such as age and extent of chest x-ray disease involvement were included in the initial multivariable model. Conditional mixed-effects model was fitted to the BMI data in view of the informative missing data (non-ignorable). Age was the only variable that was dropped from all the final multivariable models because it did not improve the model fits. The complete set of the covariates at each visit were used in the analysis; neither last- observed-carried-forward (LOCF) nor imputation was applied. Differences in rates of

FFMI, FMI, and BMI change were assessed by fitting in separate models with baseline body wasting for FFMI, FMI, and BMI cutoffs*time interaction term variables. To assess for the confounding effects of HIV, we fitted stratified models according to HIV strata.

Follow-up data were censored at the last visit. We assessed for the basic assumptions of multivariate normality and linearity, all final models had normally distributed residuals.

The analysis will be performed using SAS MIXED procedures (SAS Institute). All analyses were performed using SAS version 9.1.3 Cary software, North Carolina SAS

Institute Inc. 2004.

Results

Descriptive statistics

Of the 717 patients who were included in the analysis, 293 had reduced BMI (<18.5 kg/m2 for women and men) at diagnosis (Figure 10:1). Among the 293 with reduced

308

BMI, 153 were HIV positive and 115 were women. Two hundred seventy eight patients,

a subset of the 717 were involved in the fat and fat-free mass analysis (Figure 10:2). Of

the 278 patients, 94 had reduced FFMI (<16.7 kg/m2 for men and <14.6 kg/m2 for

women). Among the 94 with reduced FFMI, 41 were HIV positive and 19 were women.

Of the 278 patients, 120 had reduced FMI (<1.8 kg/m2 for men and <3.9 kg/m2 for

women). Among the 120 with reduced FMI, 52 were HIV positive and 69 were women.

Baseline characteristics of the study population are presented in Table 10:2. There were

gender differences at the time of diagnosis. Men had significantly higher proportions of

reduced FFMI and BMI compared to women whereas women had significantly higher

proportions of reduced FMI compared to men. There were no differences in body wasting

between HIV positive and HIV negative patients at presentation regardless of nutritional

assessment measure suggesting that HIV probably does not affect body wasting in co- infected patients as previously reported (Mupere et al. 2010; Paton and Ng 2006).

Patients who presented with history of prior smoking and moderate/or far advanced disease extent on chest x-ray had higher frequency of body wasting as assessed by FFMI and BMI compared to those without history of prior smoking and without moderate/or far advanced disease on chest x-ray (Table 10:2).

309

Rate of change for fat-free mass, fat mass, and body mass index in univariate spline

models

Fat-free mass index

Overall, the population average FFMI linear trajectory had a rate of 2.43 (95% CI, 1.34,

3.52) kg/m2 increase per month before month 3 on tuberculosis treatment in univariate mixed effects model with FFMI as the dependent variable (Table 10:3). However, after

month 3 the FFMI decreased significantly at a rate of -2.57 (95% CI; -3.87, -1.27) kg/m2

per month. There were minimal changes in FFMI after month 12 over the subsequent one

year. On average at baseline, month 3, and month 12, FFMI was -1.40 (95% CI; -1.69, -

1.11), -0.76 (95% CI; -1.17, -0.35), and -1.49 (95% CI; -2.12, -0.85) kg/m2 significantly

lower among patients that presented with reduced fat-free mass compared to patients that had normal fat-free mass, respectively. Patients that presented with reduced fat-free mass

had a significantly rapid gain in FFMI before 3 months on treatment at rate of 4.02 (95%

CI; 3.14, 4.91) kg/m2 compared to a modest gain of 1.59 (95% CI; 0.96, 2.23) kg/m2 per

month among those that had normal fat-free mass (Table 10:3). The difference in slopes

was 2.43 (95% CI; 1.34, 3.52). Before month 3 is the intensive phase of tuberculosis

treatment. There were no gains in FFMI after month 3 regardless of the amount of fat-

free mass a patient may have presented with at diagnosis. This period coincides with

continuous phase of tuberculosis treatment. However after month 12, patients presenting

with reduced fat-free mass maintained the significant rapid gain compared to patients that

had normal fat-free mass as suggested by the significant difference in slopes of 2.52

(95% CI; 1.17, 3.87) after 12 months (Table 10:3).

310

Fat mass index

In univariate mixed effects spline model for FMI (Table 10:3), there was a significant

gain in the average rate of FMI linear trajectory for the overall population before month 3

at 2.41 (95% CI: 1.33, 3.48) per month. After month 3, the rate of linear trajectory

decreased at a rate of -2.46 (95% CI: -3.78, -1.13) per month. The rate of linear trajectory was minimal per year after month 12. On average, FMI was -2.28 (95% CI; -2.76, -1.80),

-1.67 (95% CI; -2.18, -1.16), and -2.50 (95% CI; -3.37, -1.62) kg/m2 significantly lower

among patients presenting with reduced fat mass compared to patients presenting with

normal fat mass at baseline, month 3, and month 12, respectively. Patients presenting

with reduced fat mass had a significantly rapid gain in FMI before month 3 at rate of 2.99

(95% CI; 2.18, 3.80) kg/m2 compared to a no significant gain of 0.58 (95% CI; -0.12,

1.29) kg/m2 per month among those presenting with normal fat mass at diagnosis (Table

10:3). There was significant difference in slopes of 2.41 (95% CI; 1.33, 3.48). Both

patients with reduced or normal fat mass at presentation gained in FMI at the same rate

per month after month 3. However after month 12, only patients presenting with reduced

fat mass attained the significant gain per year compared to patients that presented with

normal fat mass with a difference of 2.62 (95% CI: 2.22, 4.03) in slopes (Table 10:3).

Body mass index

Unadjusted rate of BMI changes in mixed effects spline model with BMI as the

dependent variable are presented in Table 10:3. The overall population average rate of

BMI linear trajectory gain was 2.72 (95% CI, 1.74, 3.69) kg/m2 per month before month

311

3. However, after month 3 the BMI decreased significantly at a rate of -2.94 (95% CI; -

4.07, -1.81) kg/m2 per month. There was minimal change in the total population rate of

BMI change after month 12. On average at baseline and at month 12, population BMI was 2.72 (95% CI; 2.74, 3.69) and 2.23 (95% CI; 1.61, 4.05) kg/m2 significantly lower among patients that presented with reduced BMI compared to patients that presented with normal BMI, respectively. There was minimal difference in population BMI at month 3 between patients that presented with reduced BMI and normal at diagnosis. Patients that presented with reduced BMI had a significantly higher rate of gain in BMI before 3 months and after month 12 compared to the rate among patients that presented with normal BMI (Table 10:3). The differences in slopes were 2.72 (95% CI; 1.74, 3.69) before month and 2.83 (95% CI; 1.71, 4.05) after month 12, respectively. There was no difference in slope for rate of BMI gain between patients had and patients that did not have reduced BMI at presentation after month 3.

Rate of change for fat-free mass, fat mass, and body mass index in multivariable mixed spline models and stratification models according to gender

Fat-free mass

In multivariable mixed effects two spline model at month 3 and at month 12 with FFMI as the dependent variable and after adjusting for HIV, anemia (hemoglobin ≤10 mg/dl), prior smoking status, history of weight loss, and extent of disease on chest x-ray, the population average for FFMI at baseline, month 3, and month 12 was -1.60 (95% CI: -

1.90, -1.31), -1.02 (95% CI: -1.40, -0.64), and -1.10 (95% CI: -1.50, -0.70) kg/m2

312

significantly lower among patients that presented with reduced fat-free mass compared to patients that had normal fat-free mass, respectively. On average among women at

baseline, month 3, and month 12, FFMI was -1.85 (95% CI: -2.29, -1.40), -1.27 (95% CI:

-1.79, -0.76), and -1.27 (95% CI: -1.80, -0.73)] significant lower among individuals that presented with reduced fat-free mass compared to those that had normal fat-free mass at presentation, respectively. Similarly, among men at baseline, month 3, and month 12,

FFMI was -2.15 (95% CI; -2.45, -1.85), -1.68 (95% CI: -2.11, -1.24), and -1.89 (95% CI:

-2.35, -1.43) for individuals that presented with reduced fat-free mass compared to patients that had normal fat-free mass, respectively.

In the overall two spline model during the first three months, patients that presented with reduced fat-free mass had a significantly gain in FFMI at rate of 2.68 (95% CI: 0.68,

4.67) kg/m2 per month whereas patients that presented with normal fat-free mass had minimal change in FFMI at a rate of 0.35 (95% CI: -1.38, 2.08) kg/m2 per month (Table

10:4). In stratified mixed effects two spline models, the gain in FFMI among patients that

presented with reduced fat-free mass at diagnosis was however dramatic in men at rate of

4.55 (95% CI: 1.26, 7.83) (Table 10:5). The gain was minimal among women that

presented with reduced fat-free mass with a rate of 2.07 (95% CI: -0.74, 4.88). The

slopes were significantly different during the first three months between patients that

presented with reduced fat-free mass and those that had normal fat-free mass at diagnosis

for the overall population model and in stratified models for women and men. There were

no changes in FFMI after month 3 and during the one year of follow-up after month 12 in

the total population and in stratified models for women and men (Tables 10:4 and 10:5).

313

Fat mass

After adjusting for HIV, anemia (hemoglobin ≤10 mg/dl), prior smoking status, history of weight loss, and extent of disease on chest x-ray in multivariable mixed effects two spline model at month 3 and at month 12 with FMI as the dependent variable, the total population average for FMI at baseline, month 3, and month 12, was -2.15 (95% CI: -

2.62, -1.69), -1.57 (95% CI: -2.06, -1.08), and -1.64 (95% CI: -2.24, -1.04)] significant lower among patients that presented with reduced fat mass compared to those that had normal fat mass at diagnosis, respectively. On average among women at baseline, month

3, and month 12, FMI was -4.02 (95% CI: -4.65, -3.39), -3.17 (95% CI: -3.90, -2.58), and

-3.46 (95% CI: -4.34, -2.37) significant lower among individuals that presented with reduced fat mass compared to those that had normal fat mass at diagnosis, respectively.

Similarly, among men at baseline, month 3, and month 12, FMI was -1.28 (95% CI; -

1.47, -1.09), -0.76 (95% CI: -1.09, -0.44), and -0.88 (95% CI: -1.26, -0.50) for individuals that presented with reduced fat mass compared to patients that had normal fat mass, respectively.

During the first three months in the overall two spline model, patients that presented with reduced fat mass had a significantly gain in FMI at rate of 2.23 (95% CI: 0.30, 4.16) kg/m2 per month whereas patients that presented with normal fat mass had minimal

change in FMI at a rate of -0.11 (95% CI: -1.98, 1.73) kg/m2 per month (Table 10:4). In

stratified mixed effects two spline models, the gain in FMI among patients that presented

with reduced fat mass at diagnosis was higher in women at rate of 3.55 (95% CI: 0.40,

314

6.70) compared to men at rate of 3.16 (95% CI: 0.80, 5.52) (Table 10:5). The slopes were

significantly different during the first three months between patients that presented with

reduced fat mass and those that had normal fat mass at diagnosis for the overall

population model and in stratified models for women and men. After month 12 during the

one year follow-up, both patients that presented with/ or without reduced fat mass had

modest significant gains in fat mass; however, there was no significant difference in

slopes (Tables 10:4 and 10:5). There were minimal changes in FMI after month 3 and

before month 12 among women and men regardless of initial fat mass.

Body mass index

The population average BMI at baseline, month 3, and month 12 in multivariable mixed

effects two spline model at month 3 and at month 12 with BMI as the dependent variable

and after adjusting for HIV, anemia (hemoglobin ≤10 mg/dl), prior smoking status,

history of weight loss, and extent of disease on chest x-ray was -3.69 (95% CI; -4.09, -

3.30), -3.30 (95% CI; -3.71, -2.88), and -3.62 (95% CI; -4.16, -3.07) kg/m2 significantly

lower among patients that presented with reduced BMI compared to patients that had normal BMI, respectively. On average among women at baseline, month 3, and month

12, BMI was -4.23 (95% CI: -4.97, -3.48), -3.89 (95% CI: -4.67, -3.12), and -3.86 (95%

CI: -4.88, -2.84)] significant lower among individuals that presented with reduced BMI compared to those that had normal BMI at diagnosis, respectively. Similarly, among men at baseline, month 3, and month 12, BMI was -3.13 (95% CI; -3.49, -2.77), -2.71 (95%

315

CI: -3.13, -2.29), and -3.19 (95% CI: -3.71, -2.66) for individuals that presented with reduced BMI compared to individuals that had normal BMI, respectively.

In the overall two spline model during the first three months, patients that presented with

reduced BMI had a significantly gain in BMI at rate of 3.83 (95% CI: 1.67, 5.99) kg/m2

per month whereas patients that presented with normal BMI had modest gain in BMI at a

rate of 2.25 (95% CI: 0.33, 4.17) kg/m2 per month (Table 10:4). In stratified mixed

effects two spline models, the gain in BMI among patients that presented with reduced

BMI at diagnosis was however substantial in men at rate of 6.45 (95% CI: 3.02, 9.87)

(Table 10:5). The gain was minimal among women that presented with reduced BMI with

a rate of 3.30 (95% CI: -0.11, 6.72). The slopes were significantly different during the

first three months between patients that presented with reduced BMI and those that had

normal BMI at diagnosis for the overall population model and in stratified models for

men. There were minimal changes in BMI after month 3 and during the one year of

follow-up after month 12 in the total population and in stratified models for women and

men (Tables 10:4 and 10:5).

316

Rate of change for fat-free mass, fat mass, and body mass index in multivariable

mixed spline models and stratification models according to HIV status

Fat-free mass

In multivariable mixed effects two spline model at month 3 and at month 12 with FFMI

as the dependent variable when HIV was replaced by gender among the adjusters (Tables

10:6 and 10:7), the population average for FFMI at baseline, month 3, and month 12 was

-2.05 (95% CI: -2.31, -1.80), -1.57 (95% CI: -1.90, -1.24), and -1.72 (95% CI: -2.06, -

1.37) kg/m2 significantly lower among patients that presented with reduced fat-free mass compared to patients that had normal fat-free mass, respectively. On average among HIV negative individuals in stratified models at baseline, month 3, and month 12, FFMI was -

1.77 (95% CI: -2.04, -1.49), -1.35 (95% CI: -1.78, -0.93), and -1.51 (95% CI: -1.93, -

1.10)] significant lower among individuals that presented with reduced fat-free mass compared to those that had normal fat-free mass at presentation, respectively. Similarly, among HIV positive individuals at baseline, month 3, and month 12, FFMI was -2.40

(95% CI; -2.83, -1.98), -1.80 (95% CI: -2.32, -1.28), and -1.98 (95% CI: -2.51, -1.33) significantly lower for individuals that presented with reduced fat-free mass compared to individuals that had normal fat-free mass, respectively.

In the overall model and in stratified models according to HIV status during the three months, after month 3, and during the one year period of follow-up after month 12, there were minimal changes in FFMI regardless of the initial fat-free mass level at diagnosis.

However, the slopes were significantly different during the first three months between

317

patients that presented with reduced fat-free mass and those that had normal fat-free mass at diagnosis for the overall population model and in stratified models according to HIV status (Tables 10:6 and 10:7).

Fat mass

When HIV was replaced by gender among the adjusters in multivariable mixed effects two spline model at month 3 and at month 12 with FMI as the dependent variable, the total population average for FMI at baseline, month 3, and month 12, was -2.69 (95% CI:

-3.04, -2.33), -2.03 (95% CI: -2.45, -1.62), and -2.26 (95% CI: -2.74, -1.77) significant lower among patients that presented with reduced fat mass compared to those that had normal fat mass at diagnosis, respectively. On average in stratified models among HIV positive individuals at baseline, month 3, and month 12, FMI was -2.59 (95% CI: -3.06, -

2.12), -2.18 (95% CI: -2.73, -1.63), and -2.28 (95% CI: -2.09, -1.60) significant lower among individuals that presented with reduced fat mass compared to those that had normal fat mass at diagnosis, respectively. Similarly, among HIV negative indivdiuals at baseline, month 3, and month 12, FMI was -2.85 (95% CI; -3.40, -2.30), -1.77 (95% CI: -

2.36, -1.17), and -2.23 (95% CI: -2.90, -1.56) for individuals that presented with reduced

fat mass compared to patients that had normal fat mass, respectively.

During the first three months in the overall two spline model, patients that presented with

reduced fat mass had a significantly gain in FMI at rate of 2.10 (95% CI: 0.25, 3.96)

kg/m2 per month whereas patients that presented with normal fat mass had minimal

change in FMI at a rate of -0.53 (95% CI: -2.32, 1.27) kg/m2 per month (Tables 10:6 and

318

10:7). In stratified mixed effects two spline models according to HIV status, there were

minimal changes in FMI regardless of the initial fat mass level. However, the slopes were

significantly different during the first three months between patients that presented with

reduced fat mass and those that had normal fat mass at diagnosis for the overall

population model and in stratified models according to HIV status. After month 3 and

before month 12 in the overall population model and in stratified models according to

HIV status, patients that presented with normal fat mass at diagnosis, had significant gain

in FMI; however, there were no differences in slopes between the rate of increase in

patients that presented with reduced fat mass and those that had normal fat mass (Tables

10:6 and 10:7).

Body mass index

The population average BMI at baseline, month 3, and month 12 in multivariable mixed

effects two spline model at month 3 and at month 12 with BMI as the dependent variable

when HIV was replaced by gender among the adjusters was -3.59 (95% CI; -3.98, -3.20),

-3.22 (95% CI; -3.63, -2.80), and -3.47 (95% CI; -4.02, -2.03) kg/m2 significantly lower

among patients that presented with reduced BMI compared to patients that had normal

BMI, respectively. On average among HIV negative individuals in stratified models at

baseline, month 3, and month 12, BMI was -3.49 (95% CI: -4.08, -2.91), -2.91 (95% CI: -

3.56, -2.26), and -3.15 (95% CI: -3.95, -2.34) significant lower among individuals that presented with reduced BMI compared to those that had normal BMI at diagnosis, respectively. Similarly, among HIV positive individuals at baseline, month 3, and month

319

12, BMI was -3.70 (95% CI; -4.23, -3.17), -3.48 (95% CI: -4.01, -2.95), and -3.70 (95%

CI: -4.43, 2.97) for individuals that presented with reduced BMI compared to individuals

that had normal BMI, respectively.

In the overall two spline model during the first three months, patients that presented with

reduced BMI had a significantly gain in BMI at rate of 2.87 (95% CI: 0.78, 4.96) kg/m2

per month whereas patients that presented with normal BMI had minimal gain in BMI at

a rate of 1.38 (95% CI: -0.42, 3.17) kg/m2 per month (Tables 10:6 and 10:7). In stratified mixed effects two spline models, the gain in BMI among patients that presented with reduced BMI at diagnosis was however substantial in HIV positive individuals at rate of

3.16 (95% CI: 0.51, 5.82). The gain was minimal among HIV negative individuals that presented with reduced BMI with a rate of 3.39 (95% CI: -0.03, 6.80). The slopes however were not different during the first three months between patients that presented with reduced BMI and those that had normal BMI at diagnosis for HIV positive individuals in the stratified model. There were marginal changes in BMI after month 3 and during the one year of follow-up after month 12 in the total population and in stratified models for HIV negative and HIV negative individuals (Tables 10:6 and 10:7).

Discussion

To our knowledge, this is the first study to report a two year follow-up of changes in body composition and how the changes differ by gender and by HIV status among tuberculosis patients in urban Uganda, Kampala. In this retrospective cohort study of adult patients with pulmonary tuberculosis in urban Uganda, body wasting as measured

320

by reduced fat-free mass, fat mass, and body mass indexes was associated with a substantial linear increase in fat-free mass, fat mass and body mass index during the first three of months tuberculosis treatment, but the increase varied by gender and not by HIV status of the patient. Changes in body composition among men were affected by the initial fat-free mass and BMI whereas in women, body composition changes were affected by the initial fat mass. The body composition changes among HIV positive and

HIV negative patients with tuberculosis appear not to be influenced by the initial fat-free mass, fat mass, and BMI at diagnosis. There were minimal changes in body composition after three months of tuberculosis treatment and during the one year period of follow-up after month 12 among regardless of the initial body composition, gender and HIV status suggesting lack of catch-up to normal nutritional status for patients that present with body wasting.

Our findings appear to suggest that in the face of body wasting and malnutrition during the first three months of tuberculosis treatment, changes in body composition are affected by gender but not HIV. HIV appears to play a marginal role for changes in BMI in co- infected individuals. Men that had reduced fat-free mass and BMI at diagnosis attained dramatic linear increases in fat-free mass and BMI during the first three months of tuberculosis treatment compared to men that had normal fat-free mass and BMI.

However, women who presented with reduced fat mass had substantial linear increase in fat mass compared to women that had normal fat mass. There were no changes in FFMI

321

and FMI during the first three months of tuberculosis treatment regardless of HIV status.

There was no difference in slopes between patients that had reduced BMI and those that had normal BMI at diagnosis regardless of HIV status during the first three months of tuberculosis treatment.

This study is the first to report that the initial fat-free mass and BMI influences the subsequent changes in body composition among men with tuberculosis. However, the notion that the initial fat mass in women may influence subsequent changes in body composition has been reported previously in during HIV (Swanson et al. 2000). Results of the present study are consistent with previous studies (Paton et al. 2004) that revealed total lean tissue referred to as fat-free mass increased during the first six weeks of tuberculosis treatment and thereafter the increase was in fat mass. In our present study, there were minimal changes in body composition after month 3 with slight increase in fat mass for men and slight increase in BMI for HIV positive individuals. The prior study

(Paton et al. 2004); however, was limited by sample size of only 36 participants, the study population comprised of only patients who were wasted (BMI <18.5 kg/m2) and had no HIV, and it had a short duration of follow-up for only 6 months. The strengths of the present study include: large number of study participants comprising of women and men, both wasted and not wasted, both HIV positive and HIV negative individuals, and long duration of follow-up.

322

The present study has shown that patients who presented with reduced fat and fat-free mass gained in body composition at a higher rate than patients who presented with normal levels of fat and fat-free mass, particularly during the first three months of tuberculosis treatment. However, the gain was predominantly fat mass among women whereas men gained predominantly fat-free mass plus substantial levels of fat mass. The potential explanation appears to rest on the gender differences in fat-free mass content.

Women tend to have low fat-free mass compared to men, yet fat-free mass has been shown to be a significant determinant of fat oxidation (Nagy et al. 1996; Toth et al.

1996). In several reports after adjusting for difference in fat-free mass, resting fat oxidation is lower in women than men (Nagy et al. 1996; Horton et al. 1998; Toth et al.

1998). Thus, this gender differences in fat metabolism as explained elsewhere (Blaak

2001), is associated with lower basal fat oxidation that may contribute to the increased fat storage or gain in women compared to men. Moreover, longitudinal studies have shown that low rates of fat utilization predict subsequent weight (Zurlo et al. 1990). Besides differences in lipolysis, other factors such as differences in hormone action may contribute to differences in fat oxidation. Higher levels of androgens in men stimulate the formation of the nucleic acids essential for protein biosynthesis (Mooradian, Morley, and

Korenman 1987) with eventual increase in lean tissue.

In normal healthy adults, protein catabolism equals protein anabolism; however, protein catabolism exceeds protein anabolism in tuberculosis (Macallan et al. 1998). In general; however, the catch-up linear increase in lean tissue during the first three months among patients that presented with reduced fat-free mass levels suggest that patients with

323

tuberculosis can mount a protein anabolic response during treatment (Paton et al. 2003;

Macallan et al. 1998). It may be possible that the degree of body wasting offsets the

stimulus of tuberculosis on protein metabolism till a net state of anabolism (Tomkins et

al. 1983) even as the patient become sterile from the causative organism during

treatment. In the present study, there minimal changes in fat-free mass regardless of

gender and HIV status after month 3 on follow-up. Alternatively, the response could be explained by the effective adaptive response of protein metabolism to chronic inflammatory state (Paton et al. 2001).

Despite the dramatic linear increase in fat-free mass, fat mass, and BMI among patients who presented with body wasting, these patients did not normalize their indices for body composition. That is they did not regain body composition comparable to patients presenting without wasting. There may be several explanations for this. It is possible that patients who present with wasting at the time of tuberculosis diagnosis were of slim body build before disease as compared to the patients that present with normal nutritional status. While on tuberculosis treatment, the patients with apparent regain the original body composition status prior to onset of tuberculosis disease. Following attainment of the original body composition, there are minimal changes in lean tissue and individuals adapt effective energy-sparing mechanism in balance with the usual energy intake

(Kurpad, Muthayya, and Vaz 2005; Ferro-Luzzi et al. 1997). It is also possible that households where tuberculosis occurs in Uganda, and in other countries in sub-Saharan region, there is food insecurity compromising energy intake. Thus, persistently low body composition may mark lack of access to sufficient calorie and protein intake. The

324

persistently low body composition measures even after effective tuberculosis treatment

may be a marker for future health risks. Among patients with wasting who do not

normalize, there may be consequences regarding survival and physical function in the

event of any future disease insult. Further research is needed to understand the health risk

among these patients.

In this study, we used the BIA method in measurement of body composition, yet it is not

of reference standard like the dual-energy x-ray absorptiometry. The BIA prediction method used has not yet been validated in the local population. As a result, findings of body composition may be biased because of variations in hydration across ethnic groups

(Kyle et al. 2004). However, the equations that were used in this study were previously cross validated in individuals of different race (white, black, and Hispanic) among men and women, who were both healthy controls and HIV-infected patients (Kotler et al.

1996). Moreover, the equations have been used widely in other studies from Africa with meaningful findings (Shah et al. 2001; Van Lettow et al. 2004; Villamor et al. 2006;

Mupere et al. 2010). We also took care to take measurements at rest, with proper placement of leads, in participants who had not exercised or taken alcohol, in participants with voided bladder and ambient temperature. However, measurements were in patients with underlying illness that may cause shifts in body water compartments, thereby affecting measurements of fat mass. Our findings also limited by the lack of dietary intake assessment to give further insight in the interpretation of gender differences in longitudinal body composition changes.

325

Despite limitations of the present study, findings in this study revealed remarkable gender

but not HIV differences in longitudinal body composition changes during the initial phase

of treatment among tuberculosis patients that presented with body wasting and

malnutrition. Body composition changes among men were affected by the initial fat-free

mass whereas among women by fat mass. Further evaluation is needed to under the impact of providing nutritional interventions as adjuvant treatment on body composition among TB patients in sub-Saharan and evaluation of nutritional status should be involve

BMI, fat and fat-free mass.

Acknowledgements

We thank all study staff members of the Case Western Reserve University and Makerere

University research collaboration at the Tuberculosis Research Unit in the United States

and in Uganda for their assistance; the faculty of staff at Case Western Reserve

University Department of Epidemiology and Biostatistics for the guidance in analyzing

the project; and the Fogarty International Center, for the continued support.

This study was supported in part by the AIDS International Training Research Program,

Fogarty International Center, Grant No. TW000011, based at Case Western Reserve

University, Department of Epidemiology and Biostatistics, and Tuberculosis Research

326

Unit (grant N01-AI95383 and HHSN266200700022C/ N01-AI70022 from the NIAID).

This work was part of Ezekiel Mupere’s PhD thesis at Case Western Reserve University.

327

Table 10:2 Baseline characteristics of the study population with/without baseline wasting

FFMI FMI BMI

Characteristic Normal Wasting Normal Wasting Normal Wasting

(n=184) (n=94) (n=158) (n=120) (n=424) (n=293)

n (%) n (%) n (%) n (%) n (%) n (%)

Age (years)

≤30 122 (66) 54 (57) 99 (63) 77 (64) 252 (59) 170 (58)

>30 62 (34) 40 (43) 59 (37) 43 (36) 172 (41) 123 (42)

Sex

Female 115 (62) 19 (20)a 65 (41) 69 (58)b 232 (55) 115 (39)a

Male 69 (38) 75 (80) 93 (59) 51 (42) 192 (45) 178 (61)

HIV-serostatus

Negative 105 (57) 53 (56) 90 (57) 68 (57) 198 (47) 140 (48)

Positive 79 (43) 41 (44) 68 (43) 52 (43) 226 (53) 153 (52)

HGB (mg/dl)1

>10 140 (76) 72 (77) 135 (85) 77 (64)a 208 (81) 118 (64)a

328

≤10 44 (24) 22 (23) 23 (15) 43 (36) 48 (19) 67 (36)

Smoker2

No 159 (87) 56 (60)a 127 (80) 88 (74) 364 (86) 210 (72)a

Yes 24 (13) 38 (40) 31 (20) 31 (26) 58 (14) 82 (28)

Takes alcohol3

No 135 (74) 59 (63) 106 (67) 88 (74) 259 (61) 192 (66)

Yes 48 (26) 35 (37) 52 (33) 31 (26) 164 (39) 101 (34)

Extent CXR4

38 (21) 8 (9)b 26 (17) 20 (17) 74 (18) 33 (11)b Normal/minimal

Mod/advanced 143 (79) 86 (91) 131 (83) 98 (83) 346 (82) 257 (89)

Weight loss5

No 32 (17) 12 (13) 31 (20) 13 (11) 107 (25) 52 (18)b

Yes 151 (83) 82 (87) 126 (80) 107 (89) 315 (75) 240 (82) ap-value <0.001, bp-value <0.05. BMI = body mass index, FFMI = fat-free mass index, FMI = fat mass index. 1276 missed hemoglobin measurement due to lack of blood for BMI data; 4three missed extent variable in BIA data and 7 in BMI data; 3one missed history of prior smoking in BIA data and 3 missed in BMI data; 3one missed history of alcohol intake for BMI and BIA data; 5one missed history of weight loss for BIA data and 3 missed in BMI data; FFMI wasting (<16.7 kg/m2 for men and <14.6 kg/m2 for women, normal (≥ 16.7kg/m2 for men, ≥14.6 kg/m2 for women); reduced BMI <18.5 kg/m2.

329

Table 10:3 Unadjusted rate of change for fat-free mass (FFMI), fat mass (FMI), and body mass index (BMI) among pulmonary patients with reduced FFMI, FMI, and

BMI in Kampala, Uganda

Characteristic Rate SE 95% CI

Baseline fat-free mass index

Not wasted, slope before 3 mo 1.59 0.32 0.96, 2.23

Wasted, slope before 3 mo 4.02 0.45 3.14, 4.91

Not wasted, slope after 3 mo 0.06 0.11 -0.15, 0.28

Wasted, slope after 3 mo -0.07 0.16 -0.39, 0.24

Not wasted, slope after 12 mo 0.11 0.09 -0.06, 0.29

Wasted, slope after 12 mo 0.06 0.14 -0.21, 0.33

Difference in slope before 3 mo 2.43 0.56 1.34, 3.52

Difference in slope after 3 mo -0.14 0.19 -0.52, 0.24

Difference in slope after 12 mo -0.05 0.16 -0.37, 0.27

Baseline fat mass index

Not wasted, slope before 3 mo 0.58 0.36 -0.12, 1.29

Wasted, slope before 3 mo 2.99 0.41 2.18, 3.80

330

Not wasted, slope after 3 mo 1.07 0.16 0.77, 1.38

Wasted, slope after 3 mo 1.02 0.18 0.67, 1.38

Not wasted, slope after 12 mo 0.23 0.13 -0.03, 0.49

Wasted, slope after 12 mo 0.39 0.15 0.10, 0.68

Difference in slope before 3 mo 2.41 0.55 1.33, 3.48

Difference in slope after 3 mo -0.05 0.24 -0.52, 0.42

Difference in slope after 12 mo 0.17 0.20 -0.22, 0.55

Baseline body mass index

Not wasted, slope before 3 mo 3.94 0.32 3.32, 4.56

Wasted, slope before 3 mo 6.66 0.38 5.91, 7.41

Not wasted, slope after 3 mo 0.94 0.13 0.69, 1.18

Wasted, slope after 3 mo 0.72 0.15 0.42, 1.02

Not wasted, slope after 12 mo 0.30 0.10 0.10, 0.50

Wasted, slope after 12 mo 0.20 0.12 -0.05, 0.40

Difference in slope before 3 mo 2.72 0.50 1.74, 3.69

Difference in slope after 3 mo -0.22 0.20 -0.61, 0.17

Difference in slope after 12 mo -0.11 0.16 -0.42, 0.21

331

Fat-free mass wasting = FFMI <16.7 kg/m2 for men and <14.6 kg/m2 for women, fat mass wasting = FMI <1.8 kg/m2 for men and <3.9 kg/m2 for women; reduced BMI <18.5 kg/m2 for women and men.

332

Table 10:4 Adjusted rate of change for fat-free mass (FFMI), fat mass (FMI), and body mass index (BMI) among tuberculosis patients presenting with reduced FFMI,

FMI, and BMI in Kampala, Uganda

Characteristics Overall model

Rate SE 95% CI

Baseline fat-free mass index

Not wasted, slope before 3 mo 0.35 0.88 -1.38, 2.08

Wasted, slope before 3 mo 2.68 1.05 0.68, 4.67

Not wasted, slope after 3 mo 0.52 0.31 -0.09, 1.13

Wasted, slope after 3 mo 0.41 0.36 -0.30, 1.12

Not wasted, slope after 12 mo 0.02 0.23 -0.44, 0.48

Wasted, slope after 12 mo -0.05 0.28 -0.60, 0.50

Difference in slope before 3 mo 2.33 0.57 1.20, 3.45

Difference in slope after 3 mo -0.10 0.20 -0.51, 0.30

Difference in slope after 12 mo -0.07 0.17 -0.40, 0.26

Baseline fat mass index

333

Not wasted, slope before 3 mo -0.11 0.93 -1.94, 1.73

Wasted, slope before 3 mo 2.23 0.98 0.30, 4.16

Not wasted, slope after 3 mo 1.18 0.41 0.38, 1.98

Wasted, slope after 3 mo 1.08 0.43 0.23, 1.93

Not wasted, slope after 12 mo 0.38 0.31 -0.24, 0.99

Wasted, slope after 12 mo 0.58 0.34 -0.08, 1.24

Difference in slope before 3 mo 2.33 0.55 1.26, 3.41

Difference in slope after 3 mo -0.09 0.24 -0.57, 0.38

Difference in slope after 12 mo 0.20 0.20 -0.19, 0.59

Baseline body mass index

Not wasted, slope before 3 mo 2.25 0.98 0.33, 4.17

Wasted, slope before 3 mo 3.83 1.10 1.67, 5.99

Not wasted, slope after 3 mo 1.22 0.44 0.36, 2.07

Wasted, slope after 3 mo 0.79 0.49 -0.17, 1.76

Not wasted, slope after 12 mo 0.23 0.35 -0.46, 0.91

Wasted, slope after 12 mo 0.08 0.40 -0.71, 0.87

Difference in slope before 3 mo 1.58 0.60 0.40, 2.75

334

Difference in slope after 3 mo -0.43 0.28 -0.97, 0.11

Difference in slope after 12 mo -0.14 0.23 -0.60, 0.32

Fat-free mass and BMI multivariable models were adjusted for HIV, status of anemia, prior smoking status, history of weight loss, and extent of disease on chest x-ray. Fat mass multivariable model was adjusted for HIV, prior smoking status, history of weight loss, and extent of disease on chest x-ray. Fat-free mass wasting = FFMI <16.7 kg/m2 for men and <14.6 kg/m2 for women, fat mass wasting = FMI <1.8 kg/m2 for men and <3.9 kg/m2 for women; reduced BMI <18.5 kg/m2 for women and men.

335

Table 10:5 Adjusted rate of change for fat-free mass (FFMI), fat mass (FMI), and body mass index (BMI) among tuberculosis patients presenting with reduced FFMI,

FMI, and BMI stratified according to gender in Kampala, Uganda

Stratified models

Characteristics Women Men

Rate SE 95% CI Rate SE 95% CI

Baseline fat-free mass index

Not wasted, slope before 3 mo -0.21 1.03 -2.24, 1.81 2.67 1.62 -0.52, 5.85

Wasted, slope before 3 mo 2.07 1.43 -0.74, 4.88 4.55 1.67 1.26, 7.83

Not wasted, slope after 3 mo 0.49 0.38 -0.25, 2.23 0.57 0.54 -0.50, 1.63

Wasted, slope after 3 mo 0.50 0.55 -0.57, 1.58 0.28 0.57 -0.84, 1.40

Not wasted, slope after 12 mo 0.30 0.31 -0.32, 0.92 -0.25 0.37 -0.98, 0.47

Wasted, slope after 12 mo 0.13 0.49 -0.83, 1.09 -0.28 0.39 -1.05, 0.50

Difference, slope before 3 mo 2.29 0.97 0.38, 4.19 1.88 0.77 0.36, 3.39

Difference, slope after 3 mo 0.01 0.39 -0.75, 0.77 -0.29 0.26 -0.79, 0.22

Difference, slope after 12 mo -0.17 0.36 -0.87, 0.53 -0.02 0.19 -0.40, 0.35

Baseline fat mass index

336

Not wasted, slope before 3 mo 0.13 1.38 -2.58, 2.83 1.10 1.23 -1.31, 3.51

Wasted, slope before 3 mo 3.55 1.60 0.40, 6.70 3.16 1.20 0.80, 5.52

Not wasted, slope after 3 mo 1.80 0.64 0.54, 3.07 -0.09 0.46 -1.00, 0.82

Wasted, slope after 3 mo 1.41 0.75 -0.06, 2.89 -0.24 0.46 -1.14, 0.65

Not wasted, slope after 12 mo 0.11 0.48 -0.83, 1.05 0.92 0.37 0.20, 1.65

Wasted, slope after 12 mo 0.21 0.57 -0.91, 1.33 1.03 0.35 0.35, 1.72

Difference, slope before 3 mo 3.42 0.96 1.54, 5.30 2.06 0.61 0.86, 3.25

Difference, slope after 3 mo -0.39 0.44 -1.26, 0.48 -0.15 0.23 -0.61, 0.30

Difference, slope after 12 mo 0.10 0.35 -0.59, 0.80 0.11 0.19 -0.27, 0.49

Baseline body mass index

Not wasted, slope before 3 mo 1.97 1.43 -0.83, 4.77 4.75 1.66 1.49, 8.00

Wasted, slope before 3 mo 3.30 1.74 -0.11, 6.72 6.45 1.75 3.02, 9.87

Not wasted, slope after 3 mo 1.57 0.67 0.27, 2.88 0.67 0.56 -0.42, 1.76

Wasted, slope after 3 mo 1.62 0.83 -0.00, 3.24 0.03 0.58 -1.11, 1.17

Not wasted, slope after 12 mo 0.22 0.49 -0.75, 1.18 0.51 0.48 -0.43, 1.46

Wasted, slope after 12 mo -0.11 0.62 -1.33, 1.11 0.56 0.51 -0.45, 1.57

Difference, slope before 3 mo 1.33 1.08 -0.80, 3.46 1.70 0.77 0.20, 3.20

337

Difference, slope after 3 mo 0.05 0.52 -0.98, 1.07 -0.64 0.27 -1.16, -0.12

Difference, slope after 12 mo -0.32 0.40 -1.11, 0.46 0.04 0.25 -0.44, 0.53

338

Table 10:6 Adjusted rate of change for fat-free mass (FFMI), fat mass (FMI), and body mass index (BMI) among tuberculosis patients presenting with reduced FFMI,

FMI, and BMI in Kampala, Uganda

Characteristics Overall model

Rate SE 95% CI

Baseline fat-free mass index

Not wasted, slope before 3 mo -0.35 0.87 -2.05, 1.36

Wasted, slope before 3 mo 1.59 1.04 -0.46, 3.64

Not wasted, slope after 3 mo 0.34 0.31 -0.27, 0.95

Wasted, slope after 3 mo 0.15 0.38 -0.60, 0.90

Not wasted, slope after 12 mo 0.11 0.23 -0.34, 0.57

Wasted, slope after 12 mo 0.06 0.30 -0.53, 0.65

Difference in slope before 3 mo 1.93 0.60 0.75, 3.11

Difference in slope after 3 mo -0.19 0.22 -0.62, 0.23

Difference in slope after 12 mo -0.05 0.18 -0.40, 0.30

Baseline fat mass index

339

Not wasted, slope before 3 mo -0.53 0.91 -2.32, 1.27

Wasted, slope before 3 mo 2.10 0.94 0.25, 3.96

Not wasted, slope after 3 mo 1.60 0.39 0.82, 2.37

Wasted, slope after 3 mo 1.30 0.41 0.49, 2.10

Not wasted, slope after 12 mo 0.40 0.31 -0.20, 1.00

Wasted, slope after 12 mo 0.59 0.32 -0.05, 1.23

Difference in slope before 3 mo 2.63 0.56 1.53, 3.73

Difference in slope after 3 mo -0.30 0.24 -0.77, 0.17

Difference in slope after 12 mo 0.19 0.20 -0.20, 0.58

Baseline body mass index

Not wasted, slope before 3 mo 1.38 0.91 -0.42, 3.17

Wasted, slope before 3 mo 2.87 1.07 0.78, 4.96

Not wasted, slope after 3 mo 1.35 0.41 0.55, 2.16

Wasted, slope after 3 mo 1.01 0.48 0.07, 1.96

Not wasted, slope after 12 mo 0.14 0.33 -0.51, 0.78

Wasted, slope after 12 mo 0.01 0.40 -0.76, 0.79

Difference in slope before 3 mo 1.49 0.61 0.31, 2.68

340

Difference in slope after 3 mo -0.34 0.28 -0.88, 0.20

Difference in slope after 12 mo -0.12 0.24 -0.59, 0.34

Adjusted for HIV,age, status of anemia, smoking status, history of weight loss, and extent of disease on chest x-ray

341

Table 10:7 Adjusted rate of change for fat-free mass (FFMI), fat mass (FMI), and body mass index (BMI) among tuberculosis patients presenting with reduced FFMI,

FMI, and BMI stratified according to HIV in Kampala, Uganda

Stratified models

Characteristics HIV negative HIV positive

Rate SE 95% CI Rate SE 95% CI

Baseline fat-free mass index

Not wasted, slope before 3 mo -0.26 1.04 -2.31, 1.79 -0.58 1.58 -3.69, 2.53

Wasted, slope before 3 mo 1.40 1.26 -1.08, 3.88 1.84 1.89 -1.88, 5.56

Not wasted, slope after 3 mo 0.02 0.40 -0.77, 0.81 0.71 0.54 -0.35, 1.77

Wasted, slope after 3 mo -0.20 0.49 -1.16, 0.76 0.55 0.65 -0.73, 1.83

Not wasted, slope after 12 mo 0.26 0.23 -0.18, 0.71 -0.19 0.48 -1.15, 0.76

Wasted, slope after 12 mo 0.19 0.28 -0.37, 0.75 -0.25 0.66 -1.55, 1.04

Difference, slope before 3 mo 1.66 0.72 0.24, 3.07 2.42 0.99 0.46, 4.38

Difference, slope after 3 mo -0.22 0.27 -0.76, 0.32 -0.16 0.34 -0.83, 0.51

Difference, slope after 12 mo -0.07 0.16 -0.39, 0.24 -0.06 0.39 -0.82, 0.71

Baseline fat mass index

342

Not wasted, slope before 3 mo 0.33 0.98 -1.60, 2.27 -2.54 1.76 -6.00, 0.93

Wasted, slope before 3 mo 1.97 1.04 -0.07, 4.01 1.79 1.80 -1.76, 5.35

Not wasted, slope after 3 mo 1.26 0.55 0.18, 2.34 2.04 0.64 0.78, 3.30

Wasted, slope after 3 mo 1.13 0.58 -0.06, 2.27 1.42 0.65 0.14, 2.70

Not wasted, slope after 12 mo 0.49 0.41 -0.31, 1.29 -0.15 0.49 -1.11, 0.81

Wasted, slope after 12 mo 0.41 0.42 -0.42, 1.24 0.56 0.51 -0.46, 1.57

Difference, slope before 3 mo 1.63 0.57 0.51, 2.76 4.33 1.03 2.30, 6.36

Difference, slope after 3 mo -0.13 0.31 -0.73, 0.48 -0.62 0.38 -1.37, 0.13

Difference, slope after 12 mo -0.08 0.23 -0.54, 0.37 0.71 0.34 0.03, 1.38

Baseline body mass index

Not wasted, slope before 3 mo 1.06 1.51 -1.90, 4.02 2.26 1.14 0.02, 4.51

Wasted, slope before 3 mo 3.39 1.74 -0.03, 6.80 3.16 1.35 0.51, 5.82

Not wasted, slope after 3 mo 1.06 0.63 -0.18, 2.90 1.68 0.52 0.67, 2.70

Wasted, slope after 3 mo 0.74 0.72 -0.68, 2.16 1.38 0.63 0.15, 2.62

Not wasted, slope after 12 mo 0.20 0.46 -0.70, 1.10 0.15 0.43 -0.69, 0.99

Wasted, slope after 12 mo 0.11 0.54 -0.94, 1.17 0.21 0.54 -0.85, 1.28

Difference, slope before 3 mo 2.33 0.96 0.45, 4.20 0.90 0.78 -0.63, 2.43

343

Difference, slope after 3 mo -0.32 0.40 -1.09, 0.46 -0.30 0.37 -1.03, 0.43

Difference, slope after 12 mo -0.09 0.30 -0.68, 0.51 0.06 0.34 -0.60, 0.73

Adjusted for HIV,age, status of anemia, smoking status, history of weight loss, and extent of disease on chest x-ray

344

Figure 10:1 Overall Study flow diagram

Total 745 assembled from: HHC 314 +

28 Excluded: <18 years 717 Involved in analysis

Normal BMI 424 (59%) Reduced BMI 293 (41%)

HIV positive HIV positive 226 (53%) 153 (52%) HIV negative HIV negative 198 (47%) 140 (48%) Women Women 232 (55%) 115 (39%) Men 192 (45%) Men 178 (61%)

Repeated BMI measurements at: Baseline, 2, 3, 5, 6, 12, and 24 month Total observation = 4,875 Missing observation = 781

Reduced BMI <18.5 kg/m2for men and women.

345

Figure 10:2 Study flow diagram for BIA data

341 KCH with BIA data subset of 745

63 Excluded: <18 years 278 Involved in analysis

Normal FFMI 184 (66%) Reduced FFMI94 (34%)

HIV positive 79 (43%) HIV positive 41 (44%) HIV negative 105 (57%) HIV negative 53 (56%) Women 115 (62%) Women 19 (20%)

Men 69 (38%) Men 75 (80%)

Normal FMI 158 (57%) Reduced FMI 120 (43%)

HIV positive 68 (43%) HIV positive 52 (43%) HIV negative 90 (57%) Women 65 (41%) HIV negative 68 (57%) Men 93 (59%) Women 69 (58%)

Repeated FFMI or FMI measurements at: Baseline, 3, 12, and 24 month; Total observation = 1112 Missing observation = 205

FFMI wasting (<16.7 kg/m2 for men and <14.6 kg/m2 for women, normal (≥ 16.7kg/m2 for men, ≥14.6 kg/m2 for women).

346

CHAPTER 11

SUMMARY OF RESULTS AND GENERAL CONCLUSIONS

347

DISCUSSION

Introduction

Body wasting is a cardinal feature of tuberculosis and is probably one of the determinants of disease severity and outcome. However, its etiology and its management are poorly

understood, and its assessment is overlooked in research and in clinical practice. Body

wasting is loss of body mass, most notably muscle or lean body mass referred to as fat-

free mass. The simple and portable bioelectrical impedance analysis (BIA) device

provides the precise and practical approach of partitioning body weight into fat and fat-

free mass in clinical medicine. However, it may not be easily affordable and accessible in

sub-Saharan Africa where it is needed to detect body wasting early enough among tuberculosis and HIV patients. Sub-Saharan Africa has a high burden of tuberculosis and

HIV-associated tuberculosis. The widely used body mass index (BMI) in assessing body wasting is popularly a measure of body fat and it gives unreliable results of wasting when

it is low. Thus, BMI overestimates body wasting. BMI can be partitioned into body fat

mass (FMI) and fat-free mass (FFMI) indices by dividing fat and fat-free mass with

height. Body composition in this dissertation refers to fat and fat-free mass compartments

of the body weight.

The data that was used to generate results in this dissertation were obtained from cross-

sectional and retrospective cohort designs. The cross-sectional design enrolled 131 HIV positive and HIV negative adults with or without tuberculosis in urban Uganda. In this

348

design, participants were assessed for waist circumference (WC), BMI, mid-upper-arm

circumference (MUAC), and the 24-hour dietary intake recall. In a retrospective cohort

design, 745 adult tuberculosis patients that were merged from three completed NIH

funded prospective studies at the Uganda-Case Western Reserve University research collaboration were analyzed. All patients were screened for HIV infection and had BMI

measurements at month 0, 2, 3, 5, 6, 12, and 24 on follow-up. A subset (314) of the 745

had BIA fat and fat-free mass measurements at month 0, 3, 12, and 24 on follow-up. Any case of deaths and date was documented on follow-up.

Purpose

The purpose of this dissertation was to generate information that could be used to improve survival in tuberculosis patients. This dissertation focused first on understanding simple and inexpensive approaches of assessing body wasting. Second, focused on generating information that could aid in understanding of body wasting and its management, that is whether nutritional factors influence body composition in the face of tuberculosis; and whether body wasting as measured by precise measures of body composition modifies the course of tuberculosis.

349

SUMMARY OF RESULTS

Chapter Four: Body Composition Measured with Bioelectrical Impedance Analysis and Anthropometry among HIV Positive and HV Negative Adults with or without

Tuberculosis in Urban Kampala, Uganda

Analysis in this chapter established whether known existing equations that estimated body composition among Caucasians could provide comparable results of body composition to those of BIA device in an African population in Uganda. The equations involved simple measurements such as WC, BMI, and MUAC. Findings among 131 HIV positive and HIV negative adults with or with tuberculosis revealed that existing equations that provided comparable results of body composition to those of BIA differed by gender and tuberculosis status. The equation that involved MUAC or BMI provided comparable results of lean tissue to that of BIA whereas the equation that involved

MUAC provided comparable results of fat mass among men and women with tuberculosis regardless of HIV status. Among men without tuberculosis, the equation that involved MUAC provided comparable results of fat and lean tissue to that of BIA whereas among women without tuberculosis, the equation that involved WC provided comparable results of fat and lean tissue regardless of HIV status.

350

Chapter Five: Indicators of Dietary Adequacy among HIV Positive and HIV

Negative Adults in Urban Kampala, Uganda

Analysis in this chapter involved 131 HIV positive and HIV negative adults with or

without tuberculosis from urban Uganda to assess nutritional adequacy of dietary intake and the validity of simple low-cost methods to evaluate nutritional adequacy of diets consumed among HIV positive and HIV negative adults with or without tuberculosis in urban Uganda. Findings revealed that all participants (100%) consumed at least cereals,

roots, and tubers, and 90% consumed vegetables not rich in vitamin A such tomatoes and

onions while only 45% consumed vitamin-A-rich fruits and vegetables, and only 15%

consumed eggs.

The mean FVS and DDS for the study population were low 8.1 ± 2.8 and 4.7 ± 1.4,

respectively. Both men and women regardless of tuberculosis and HIV status, had

carbohydrate and ascorbic acid deficiency in the range of 0 to 30% whereas other nutrient

intakes including energy, protein, dietary fiber, calcium, magnesium, zinc, iron, vitamin

A, vitamin D, and folate had deficiencies ranging 25% to 100%. When a MAR of 65%

was used as a cut-off point for nutrient adequacy, it was found that FVS must be 9 or

more and DDS must at least 5. Among women, both FVS and DDS had a high ability to

identify participants with an inadequate or adequate diet while among men FVS had a

high ability to identify individuals with inadequate diet but low ability to identify those

with adequate diet, DDS had low ability to identify individuals with inadequate diet but

had a high ability to identify those with adequate diet.

351

Chapter Six: Predictors of Fat and Lean Tissue among HIV Positive and HIV

Negative Adults with or without Tuberculosis in Urban Kampala, Uganda.

Analysis in this chapter aimed to establish factors that may influence fat mass, fat-free

mass (lean tissue), and BMI among 131 HIV positive and HIV negative adults with or

without tuberculosis in urban Uganda. We found that energy intake was associated with

an increase in BMI among women (0.003 ± 0.001 SE, p=0.028) although intake in the

presence of tuberculosis was associated with a decrease in BMI (-0.004 ± 0.001 SE,

p=0.008). Protein intake among women with no income (-0.02 ± 0.01 SE, p=0.027) and

among unemployed women (-0.08 ± 0.03 SE, p=0.010) was associated with a decrease in

lean tissue and fat mass, respectively whereas protein intake among women with

tuberculosis was associated with an increase in BMI (0.10 ± 0.05 SE, p=0.039). Being a

single woman (0.58 ± 0.25 SE, p=0.022) was associated with an increase in lean tissue

whereas having reduced appetite (-0.79 ± 0.33 SE, p=0.020) was associated with a

decrease in lean tissue and fat mass (-4.17 ± 1.94 SE, p=0.036). Among men, tuberculosis

(-1.42 ± 0.45 SE, p=0.003) was associated with a decrease in lean tissue. Similarly, having reduced appetite was associated with decrease in fat mass (-1.90 ± 0.90 SE, p=0.040) and BMI (-2.95 ± 0.68 SE, p<0.001). HIV did not influence body composition regardless of gender.

352

Chapter Seven: Body Wasting and Dietary Intake among HIV Positive and HIV

Negative Adults with or without Tuberculosis in Urban Kampala, Uganda

The analysis in this chapter involved 131 HIV positive and HIV negative adults with or with tuberculosis to evaluate the independent effects of tuberculosis and HIV, to evaluate whether dietary intake differs by body wasting and severity of tuberculosis disease. We found that tuberculosis patients that had moderate/or severe clinical disease had lower dietary intake for energy, protein, total fat, carbohydrate, calcium, vitamin A, and folate compared to patients with mild disease. Both men and women had comparable dietary intake among patients with TB regardless of HIV status whereas HIV negative women had reduced energy, protein, and folate intake among individuals without TB compared to men. Tuberculosis patients with wasting of lean tissue or those with reduced BMI had comparable nutrient intake with counterparts that had normal lean tissue or normal BMI.

Chapter Eight: Correlates of Dietary Adequacy among HIV Positive and HIV

Negative Adults with or without Tuberculosis in Urban Kampala, Uganda

In this chapter, we evaluated dietary correlates of energy and protein intake and correlates of inadequate nutrient intake. There was female gender interaction between having tuberculosis and reduced appetite, and between having tuberculosis and being a current alcohol taker in the model for energy intake. Women that had tuberculosis with reduced appetite or tuberculosis with history of taking alcohol had decreased energy intake. Also women who had history of alcohol intake had decreased protein intake. There was no compromise with energy and protein intake among men. Women were associated with

353

inadequate iron intake. Further, women with tuberculosis were associated with inadequate folate intake. Individuals with tuberculosis residing in households of more than two people or those with no or low education were associated with inadequate vitamin A intake.

Chapter Nine: Impact of Body Wasting on Survival among Adult Patients with

Pulmonary Tuberculosis in Urban Kampala, Uganda

In this chapter, we evaluated the impact of HIV and body wasting on survival in

tuberculosis patients using precise measures of nutritional status, the height-normalized

fat-free mass (FFMI) and fat mass (FMI) indices. During the follow-up period, 19% of

310 patients with baseline wasting by BMI died compared to 11% of 437 without

wasting, a crude risk ratio of 1.74 (95% confidence interval (CI): 1.22, 2.48). Of 103 with

baseline wasting by fat-free mass index, 16% died, compared to 7% without wasting,

crude risk ratio of 2.31 (95% CI: 1.10, 3.92).

In stratified survival analysis, survival proportion was significantly lower among men

with reduced BMI compared to men with normal BMI; and lower among women with

reduced fat-free mass index compared to women with normal fat-free mass index. In

multivariable Cox regression model using anthropometric data, the relative hazard of

death when patient had reduced BMI was 1.85 (95% CI: 1.25, 2.73). In a nested model,

the relative hazard for death was 1.70 (95% CI: 1.03, 2.81) for men with reduced BMI

and 1.83 (95% CI: 0.96, 3.50) for women with reduced BMI. In a model using fat-free

354

mass index data, the relative hazard of death when patients had reduced fat-free mass index was 1.88 (0.96, 3.65). In a nested model, the relative hazard of death was 6.83

Chapter Ten: Longitudinal Changes in Body Composition among Tuberculosis

Patients with or without Body Wasting in Urban Kampala, Uganda

In this chapter, longitudinal data analysis was conducted to assess whether body wasting as measured by height-normalized measures of nutritional status at diagnosis of tuberculosis, and whether HIV infection modifies longitudinal changes in body

composition during and after tuberculosis treatment. Results in this chapter revealed that

there were no differences in body wasting as assessed by reduced lean tissue, fat mass,

and BMI between HIV positive and HIV negative patients at diagnosis of tuberculosis. In

stratified mixed effects two spline models during the first three months of treatment, the

gain in lean tissue among patients that presented with wasted lean tissue at diagnosis was

dramatic in men with rate of 4.55 kg/m2 (95% confidence interval (CI): 1.26, 7.83) per

355

month; however, the gain was minimal among women who presented with reduced lean

tissue with rate of 2.07 kg/m2 (95% CI: -0.74, 4.88).

In stratified models for fat mass as dependent variable, women with reduced fat mass at presentation had a substantial gain in fat mass at rate of 3.55 kg/m2 (95% CI: 0.40, 6.70)

whereas men had a rate of 3.16 kg/m2 (0.80, 5.52). In stratified models with BMI as

dependent variable, men with reduced BMI at presentation gained BMI at a rate of 6.45

kg/m2 (95% CI: 3.02, 9.87) whereas women at a rate of 3.30 kg/m2 (95% CI: -0.11, 6.72).

There were minimal changes in lean tissue, fat mass, and BMI during the first three

months of treatment in stratified models according to HIV status. Further, there were

minimal changes in lean tissue, fat mass, and BMI after month 3 and during the one year

follow-up after month 12.

Conclusion

In this dissertation, we have shown that body composition can reliably be assessed using

inexpensive and easy to measure anthropometric assessments such as waist

circumference, MUAC, and BMI. Precise estimation of fat-free mass is essential to

identify patients with body wasting so that appropriate and early interventions are

instituted. The cross-sectional nature of the study limits making conclusions beyond the

time of diagnosis among tuberculosis. However, since the study population included

participants with and without tuberculosis, one can infer that the equations evaluated in

356

this dissertation can reliably be used to monitor changes in body composition during and after tuberculosis treatment.

Several conclusions were generated following our working model in this dissertation

(Figure 1). This dissertation has shown that the dietary consumption in the study population of HIV positive and HIV negative adults with or without tuberculosis from urban Uganda was monotonous, rich in carbohydrates and deficient in nutrients regardless of gender, tuberculosis, and HIV status. The ability of FVS and DDS indices to identify individuals with inadequate or adequate diet consumption differed by gender.

The FVS was a better predictor of nutritional adequacy among women whereas DDS was a better predictor among men.

Findings in this dissertation suggest that there are remarkable gender differences in how energy and protein intake influence body composition, and we found important interactions in the face of tuberculosis, and when there is no income. HIV does not appear to influence nutrient intake on body composition. At the time of tuberculosis diagnosis, the 24-hour dietary intake recall varied by severity of tuberculosis disease, but not tuberculosis disease or HIV status. However, in the absence of tuberculosis, dietary intake varied by gender. Both men and women had comparable dietary intake among patients with tuberculosis regardless of HIV status whereas HIV negative women had reduced energy, protein, and folate intake among individuals without tuberculosis compared to men. As regards dietary correlates, findings suggest that correlates of energy

357

and protein intake differ by gender. Women and individuals having tuberculosis who reside in households with two or more people or who have no or low education are at vulnerable state of inadequate nutrient intake.

We have shown in this dissertation that body wasting is associated with poor survival and this effect differed by gender. Men with reduced BMI at the time of tuberculosis had poor survival whereas women with reduced lean tissue had poor survival suggesting that BMI is a better predictor of death among men whereas wasting of lean tissue is a better predictor of death among women. Further, survival differed by gender; men had poor survival compared to women. However, this gender difference in survival was modified by HIV. It is HIV positive men that had poor survival compared to HIV positive women.

Our results suggest that body wasting at the time of tuberculosis diagnosis modifies longitudinal changes in body composition during tuberculosis treatment. The effect however, differed by gender. Men with wasted lean tissue and reduced BMI had dramatic increase in lean tissue and BMI whereas women with reduced fat mass had significant increase in fat mass during the first three months of tuberculosis treatment. HIV infection did not influence changes in body composition. Of note, there were minimal changes in body composition after month 3 and during the one year follow-up after month 12 regardless of gender, HIV status, and the initial level of body composition.

358

Figure 11:1 Overall conclusions: Tuberculosis – HIV – malnutrition model

Tuberculosis HIV Reduced appetite Immune activation Inadequate dietary intake Immune deficiency Altered metabolism: HIV replication Differ by gender Increased cytokine production Anabolic block HIV-TB Co-infection Increased energy expenditure Malabsorption Increased inadequate intake Differ by gender

Lipid and protein metabolism: Differ by gender Wasting of body composition: Differ by gender Men (Mupere et al. 2010) Women Begin with high fat-free mass Begin with low fat-free mass Begin low with fat mass Begin with high fat mass Marked fat-free mass wasting Preserve fat-free mass wasting Fat mass wasted in proportion with fat-free mass. Marked fat mass wasting Low body mass index High body mass index Gender differences in micronutrient deficiency not known?

Wasting modifies: Differ by gender Men Women Survival: Poor with low BMI Survival: Poor with low FFMI Poor HIV positive Minimal HIV positive HIV negative same as women HIV negative same as men Changes: Increase in FFM and BMI Changes: Significant increase FMI No catch-up after month 3 No catch-up after moth 3 Mechanism not known Mechanism not known Dietary intake: Differ by disease severity, not TB or HIV status or gender Predictors body composition: Energy and protein intake, TB, TB and appetite appetite, income, employment, single.

FFMI = fat-free mass index, FMI = fat mass index, BMI = body mass index.

359

Limitations and Future Studies

Findings in this dissertation, demonstrated poor survival among patients that presented with body wasting. The survival differed by gender. Men had poor survival when BMI was low, whereas women had poor survival when lean tissue was wasted. This provides compelling evidence to evaluate nutritional interventions that can improve survival among patients with body wasting. In evaluating nutritional interventions however, gender differences will need to be considered.

Our results revealed that there were minimal changes in body composition after month 3 and during the one year follow-up after month 12 regardless of gender, HIV status, or the initial level of body composition. Yet, patients that presented with wasting had dramatic increase in lean tissue and BMI for men and fat mass for women; however, there was no catch-up to normal nutritional status. Studies are needed to evaluate the mechanisms for the lack of catch-up among patients that presented with wasting and the lack of changes in body composition after month 3.

To address the limitations in this dissertation, validation studies with reference methods such as air-displacement plethysmography, underwater weighing and dual X-ray absorptiometry (DXA) are needed to validate the BIA regression equations and the existing equations that involve simple anthropometric measurements used in estimation of body composition. Validation in follow-up studies and in different populations because of variations in hydration is needed.

360

In this dissertation, a single 24-hour recall was used to assess dietary intake. However,

four multiple-pass 24-hour recalls have been shown to be the most appropriate method

for a study of diet and nutrition in low-income households (Vucic et al. 2009). A whole

year-round study is recommended to understand the habitual consumption in Uganda.

Further, evaluation is needed to understand changes in dietary intake over time and how

these changes would affect body composition and survival.

We found simple counting of food items or food groups can give fairly good assessment

of nutritional adequacy. The study however, was cross-sectional and was conducted in an

urban setting. The conclusions might therefore be valid for similar urban settings, but the

methods and approaches are equally valid for rural areas. To attain generalizability of conclusions and to understand whether dietary adequacy would change with habitual consumption, additional studies are recommended in rural settings, in different socioeconomic, and in different ethnic groups. Also follow-up studies are needed to understand how dietary adequacy changes during different seasons of the year.

Strengths

Findings in the present dissertation were generated using a large database that provided adequate power to detect differences in survival and changes in body composition by gender and body wasting. The data provided adequate power to interactions with survival between gender and HIV infection. To our knowledge, this is the first study to show the effect of wasting on survival using height-normalized fat-free mass rather than BMI that

361

may overestimate wasting among women and thus, overestimate survival due to wasting.

Height-normalized fat-free mass provide a precise measure of nutritional status than BMI

(Kyle, Piccoli, and Pichard 2003; Kyle, Genton, and Pichard 2002; VanItallie et al.

1990). The median follow-up period for participants in the data used for analysis in this

dissertation was more than two years; suggesting a longer period of follow-up than the

previous studies (Sani et al. 2006; Lucas et al. 1993). Thus, our findings are embedded with a temporal relationship between enrolment to time of death or censoring and between enrolments to different time points when changes in body composition were assessed on follow-up.

We had a full panel of control groups that enabled to this dissertation to show the

independent effects of tuberculosis and HIV on dietary intake. Further, the dissertation

has shown the biologic and socioeconomic plausibility.

Public Health Implications

Findings in this dissertation suggest the need for national tuberculosis programs to address the nutritional needs of tuberculosis patients at the time of diagnosis, during and after treatment. The nutritional needs may range from health education to provision of

nutritional supplements. Results revealed that patients with body wasting had poor

survival and those that survived after month 3, did not catch-up to normal nutritional

status. The findings provide theoretical framework for targeted nutritional intervention to

tuberculosis patients that present with body wasting, particularly women with reduced

362

lean tissue, men with reduced BMI or HIV positive men. There is potential that if

nutritional needs are addressed, this may improve survival and adherence to treatment.

Our findings contribute to an increased understanding of the potential methods of assessing body wasting using simple and inexpensive anthropometric measurements. This implies that during regular clinical practice and in public field surveys, body composition can easily be assessed to identify individuals and population groups at risk of poor outcome. Early identification of body wasting enables timely institution of management to prevent poor outcome.

The monotonous diet with substantial nutrient deficiencies in this dissertation, suggests the need for fortification of common food items in Uganda. Fortification of staple foods has been shown to improve micronutrient intake among adults in South Africa (Steyn et al. 2008). In this dissertation, we have shown that simple indices of food variety score and dietary diversity score can be used to identify individuals or families with inadequate nutrient intake. Thus, these simple scores can be used in management of tuberculosis and in public health settings as tools to provide health education so that individuals or families understand when to have an adequate diet.

To address the nutritional needs of tuberculosis patients and the nutritional needs of the community where tuberculosis patients reside, will require a collaborative effort of

363

several stakeholders including clinicians, healthcare professionals, policy makers at the

Ministry of Health in Uganda, program managers for the poverty eradication scheme and for various existing non-government organizations and the will of the political organs in the country. The clinicians, the healthcare professionals and the public health practitioners should strive to understand the reasons for poor survival, failure to attain normal nutritional status after therapy, and poor diet consumption among tuberculosis patients in order to develop and implement targeted interventions.

364

CHAPTER 12

EXTRA TABLES AND FIGURES FOR CHAPTER FOUR

365

Table 12:12:1 Mean intra-supervisor, mean intra-observer, and mean inter- observer technical error of measurement prior to data collection

Difference Technical Error of Measurement between Measurement (n=6) trainee and Intra- Intra- Inter- supervisor Supervisor Observer Observer Mean Mean Mean

Weight (kg) 0.79 0.000 0.068 0.620

Height (cm) 0.19 0.037 0.059 0.037

MUAC (cm) 0.65 0.022 0.017 0.042

Triceps ST (mm) 1.11 0.036 0.054 1.242

Biceps ST (mm) 0.39 0.029 0.049 0.155

Subscapular ST (mm) 0.30 0.034 0.057 0.088

Sacral iliac ST (mm) 0.58 0.034 0.041 0.335

Waist circumference (cm) 0.07 0.067 0.035 0.005

Hip circumference (cm) 0.43 0.043 0.026 0.184

Resistance (ohms) 0.84 0.146 0.165 0.712

Reactance (ohms) 0.23 0.044 0.046 0.051

ST = skinfold thickness, MUAC=mid-upper arm circumference.

366

Table 12:12:2 Coefficient of reliability prior to data collection

Reference(Frisancho A.R 1990;

Lohman T.G, Roche A.F, and Measurement (n=6) Martorell R 1988) Coefficient of Reliability Intra-observer Inter-observer Weight (kg) 1.200 1.500 0.99

Height (cm) 0.692 0.953 0.90

MUAC (cm) 0.347 0.425 1.00

Triceps ST (mm) 0.800 1.890 1.00

Biceps ST (mm) 0.600 0.600 0.98

Subscapular ST (mm) 1.830 1.530 0.96

Sacral iliac ST (mm) 1.000 1.700 1.00

Waist circumference (cm) 1.000 1.000 0.74

Hip circumference (cm) 1.230 1.380 1.00

Resistance (ohms) - - 0.96

Reactance (ohms) - - 0.96

ST = skinfold thickness, MUAC=mid-upper arm circumference.

367

Table 12:12:3 Guide for comparison of inter-observer error (Frisancho A.R 1990)

Measurement Difference between trainee and supervisor

Good Fair Poor Gross Error Weight (kg) 0 to 0.1 > 0.2 0.3 to 0.4 ≥ 0.5

Height (cm) 0 to 0.5 0.6 to 0.9 1.0 to 1.9 ≥ 2.0

Mid upper arm circumference (mm) 0 to 5 6 to 9 10 to 19 ≥ 20

368

Figure 12:1 Fat-free mass measured by BIA compared with fat-free mass measured by equation involving waist circumference

BIA = bioelectrical impedance analysis, FFM = fat-free mass

369

Figure 12:2 Fat mass measured by BIA compared with fat mass measured by equation involving waist circumference

BIA = bioelectrical impedance analysis, FM = fat mass, Waist = waist circumference

370

Figure 12:3 Fat-free mass measured by BIA compared with fat-free mass measured by Durnin & Womersely equations involving sum of 4 skinfold thickness

BIA = bioelectrical impedance analysis, FFM = fat-free mass, ST = skinfold thickness.

371

Figure 12:4 Fat mass measured by BIA compared with fat mass measured by

Durnin & Womersely equations that involves sum of 4 skinfold thickness

BIA = bioelectrical impedance analysis, FM = fat mass, ST = skinfold thickness.

372

Figure 12:5 Fat-free mass measured by BIA compared with fat-free mass measured by equation involving BMI

BIA = bioelectrical impedance analysis, FFM = fat-free mass, BMI = body mass index.

373

Figure 12:6 Fat mass measured by BIA compared with fat mass measured by equation involving BMI

BIA = bioelectrical impedance analysis, FM = fat mass, BMI = body mass index

374

Figure 12:7 Fat-free mass measured by BIA compared with fat-free mass measured

by equation involving MUAC

BIA = bioelectrical impedance analysis, FFM = fat-free mass, MUAC = mid-upper-arm circumference

375

Figure 12:8 Fat mass measured by BIA compared with fat mass measured by

equation involving MUAC

BIA = bioelectrical impedance analysis, FM = fat mass, MUAC=mid-upper arm circumference.

376

CHAPTER 13

EXTRA TABLES AND FIGURES FOR CHAPTER NINE

377

Table 13:1 Comparison of key baseline variables across phase II prednisolone study, household contact study, and Kawempe community health study

Household Kawempe Phase II 1Characteristics Contact study Prednisolone

[mean, (SD)] (n=95) (n=312) (n=340) p-value

Sex

Male (%) 59 (62) 163 (52) 178 (51) 0.172

Female (%) 36 (38) 151 (48) 166 (49)

HIV status

Negative (%) 0 (0) 159 (51) 186 (55) <0.001

Positive (%) 95 (100) 153 (49) 153 (45)

Age in years 30.7 (7.2) 31.8 (9.0) 29.6 (9.2) 0.003

Weight in kg 52.0 (6.8) 52.3 (8.7) 51.6 (8.4) 0.297

Height in cm 165.7 (8.8) 164.2 (9.3) 163.8 (8.7) 0.077

BMI in (kg/m2) 19.0 (2.6) 19.4 (2.9) 19.2 (2.8) 0.383

Hemoglobin in g/dl 11.0 (1.8) 11.6 (2.4) 11.4 (2.2) 0.070

Chest x-ray disease

378

Normal/mild (%) 4 (4) 54 (17) 53 (16) 0.007

Moderate/far advanced (%) 89 (96) 257 (83) 282 (84)

1Continuous variables are means ± standard deviation (SD).

379

Table 13:2 Assessing the proportional hazards assumption with a statistical test:

Correlations between ranked failure time and Schoenfeld residuals

Characteristic Mean Residuals Ranked failure time p-value

Sex -1.02x10-8 0.30 0.120

Age (years) 8.47x10-8 -0.21 0.284

HIV status 4.84x10-7 0.29 0.129

Hemoglobin (g/dl) 1.07x10-7 0.06 0.760

Fat mass index (kg/m2) -1.96x10-8 -0.05 0.788

Fat-free mass index (kg/m2) -2.16x10-8 -0.09 0.646

Body mass index (kg/m2) 2.06x10-8 -0.03 0.864

Chest x-ray disease extent -7.89x10-8 0.15 0.440

Smoker 1.75x10-8 0.06 0.766

Takes alcohol 6.80x10-8 0.24 0.215

History weight loss 5.50x10-8 -0.21 0.285

380

Table 13:3 Baseline characteristics of pulmonary tuberculosis patients between HIV negative and HIV positive

HIV negative HIV positive Characteristic (n=345) (n=401)

Sex

Females [n (%)] 165/351 (47) 186/351 (53)

Males [n (%)] 180/395 (46) 215/395 (54)

Age (years)

≤30 [n (%)] 233/434 (54) 201/434 (46)a

>30 [n (%)] 112/312 (36) 200/312 (64)

Hemoglobin (g/dl)1

>10 [n (%)] 169/342 (49) 173/342 (51)a

≤10 [n (%)] 32/121 (26) 89/121 (74)

Body mass index (kg/m2)

Normal [n (%)] 202/436 (46) 234/436 (54)

Low [n (%)] 143/310 (46) 167/310 (54)

Fat-free mass index (kg/m2)

381

Normal [n (%)] 116/207 (56) 91/207 (44)

Low [n (%)] 56/103 (54) 47/103 (46)

Fat mass index (kg/m2)

Normal [n (%)] 98/176 (56) 78/176 (44)

Low [n (%)] 71/127 (56) 56/127 (44)

Chest x-ray disease extent 46/111 (41) 65/111 (59) Normal/mild [n (%)]

Moderate/far advanced [n (%)] 296/627 (47) 331/627 (53)

Smoker

No [n (%)] 272/592 (46) 320/592 (54)

Yes [n (%)] 73/151 (48) 78/151 (52)

Currently takes alcohol

No [n (%)] 240/466 (51) 226/466 (49)a

Yes [n (%)] 105/279 (38) 174/279 (62)

History weight loss

No [n (%)] 60/165 (36) 105/165 (64)b

Yes [n (%)] 283/578 (49) 295/578 (51)

382

ap-value <0.001, bp-value <0.05.

383

Table 13:4 Multivariable Relative hazards [HR, 95% confidence intervals (CIs)] for death among tuberculosis patients with normal compared with patients having low body mass index (BMI)

Deaths/N (%) Relative 95% Confidence

Characteristic hazard Interval

Body mass index (kg/m2)

Normal 47/437 (11) 1 -

Low 58/310 (19) 13.39 1.23, 146.19

Sex

Female 38/352 (11) 1 -

Male 67/395 (17) 1.52 0.83, 2.76

Age (years)

≤30 34/434 (8) 1 -

>30 71/313 (23) 3.06 1.61, 5.83

HIV-serostatus1

Negative 6/345 (2) 1 -

Positive 99/401 (25) 42.68 5.88, 309.73

Smoker3

384

No 84/593 (14) 1 -

Yes 20/151 (13) 0.57 0.23, 1.40

Chest x-ray extent6

Normal/minimal 19/111 (17) 1 -

Moderate/far advanced 86/628 (14) 0.96 0.46, 1.99

Weight loss5

No 11/165 (7) 1 -

Yes 94/579 (16) 3.54 1.88, 6.66

Reduced BMI*sex 1.06 0.45, 2.48

Reduced BMI*HIV 0.25 0.03, 2.23

Reduced BMI*age 0.55 0.24, 1.29

Reduced BMI*smoker 1.05 0.35, 3.11

Reduced BMI*extent 0.64 0.24, 1.75

-2LL reduced model - -2LL full model: 1202.115 – 1193.495 = 8.62; df = 5, p=0.125. 1One missed HIV status; 6eight missed extent variable; 3three missed history of ever smoked; 5four missed history of weight loss; low BMI <18.5 kg/m2.

385

Table 13:5 Multivariable Relative hazards [HR, 95% confidence intervals (CIs)] for death among tuberculosis patients with normal compared with patients having low body mass index (BMI)

Deaths/N (%) Relative 95% Confidence

Characteristic hazard Interval

Body mass index (kg/m2)

Normal 47/437 (11) 1 -

Low 58/310 (19) 9.21 0.99, 85.90

Sex

Female 38/352 (11) 1 -

Male 67/395 (17) 1.52 0.85, 2.73

Age (years)

≤30 34/434 (8) 1 -

>30 71/313 (23) 3.04 1.60, 5.78

HIV-serostatus1

Negative 6/345 (2) 1 -

Positive 99/401 (25) 41.95 5.78, 304.28

Smoker3

386

No 84/593 (14) 1 -

Yes 20/151 (13) 0.60 0.36, 1.00

Chest x-ray extent6

Normal/minimal 19/111 (17) 1 -

Moderate/far advanced 86/628 (14) 0.77 0.47, 1.27

Weight loss5

No 11/165 (7) 1 -

Yes 94/579 (16) 3.53 1.88, 6.65

Reduced BMI*sex 1.03 0.45, 2.34

Reduced BMI*HIV 0.26 0.03, 2.29

Reduced BMI*age 0.56 0.24, 1.32

-2LL reduced model - -2LL full model: 1202.115 – 1194.245 = 7.87; df = 3, p=0.0488. 1One missed HIV status; 6eight missed extent variable; 3three missed history of ever smoked; 5four missed history of weight loss; low BMI <18.5 kg/m2.

387

Table 13:6 Relative hazards [HR, 95% confidence intervals (CIs)] for death among tuberculosis patients with normal compared with patients having low body mass index (BMI)

Overall model Characteristics Deaths/N (%)

HR (95% CI)

BMI (kg/m2)

Normal (≥18.5) 47/437 (11) 1

Low (<18.5) 58/310 (19) 1.78 (1.20, 2.64)

HIV status

Negative 6/345 (2) 1

Positive 99/401 (25) 18.18 (7.96, 41.52)

Gender

Female 38/352 (11) 1

Male 67/395 (17) 1.70 (1.12, 2.59)

Smoker2

388

No 84/593 (14) 1

Yes 20/151 (13) 0.62 (0.37, 1.04)

Weight loss3

No 11/165 (7) 1

Yes 94/579 (16) 3.70 (1.97, 6.94)

Chest x-ray extent4

Normal/minimal 19/111 (17) 1

Moderate/far advanced 86/628 (14) 0.73 (0.44, 1.21)

1One missed HIV status; 4eight missed extent variable; 2three missed history of ever smoked; 3four missed history of weight loss; low BMI <18.5 kg/m2.

389

Table 13:7 Relative hazards [HR, 95% confidence intervals (CIs)] for death among tuberculosis patients with normal compared with patients having low body mass index (BMI) stratified according to Age group

Stratified models

Deaths/N Young ≤30 yr Old >30 yr Characteristics (%) (n=434) (n=313)

HR (95% CI) HR (95% CI)

BMI (kg/m2)

Normal (≥18.5) 47/437 (11) 1 1

Low (<18.5) 58/310 (19) 2.85 (1.41, 5.77) 1.39 (0.86, 2.24)

HIV status

Negative 6/345 (2) 1 1

Positive 99/401 (25) 23.94 (5.70, 100.60) 11.75 (4.27, 32.36)

Gender

Female 38/352 (11) 1 1

Male 67/395 (17) 1.15 (0.57, 2.31) 1.88 (1.08, 3.25)

Smoker2

390

No 84/593 (14) 1 1

Yes 20/151 (13) 0.49 (0.17, 1.43) 0.61 (0.34, 1.11)

Weight loss3

No 11/165 (7) 1 1

Yes 94/579 (16) 1.96 (0.84, 4.56) 6.24 (2.27, 17.18)

Chest x-ray extent4

Normal/minimal 19/111 (17) 1 1

Moderate/far advanced 86/628 (14) 0.65 (0.26, 1.61) 0.78 (0.42, 1.43)

1One missed HIV status; 4eight missed extent variable; 2three missed history of ever smoked; 3four missed history of weight loss; low BMI <18.5 kg/m2.

391

Table 13:8 Relative hazards [HR, 95% confidence intervals (CIs)] for death among

HIV positive tuberculosis patients compared with HIV negative patients

Overall model Characteristics Deaths/N (%)

HR (95% CI)

HIV status

Negative 6/345 (2) 1

Positive 99/401 (25) 15.66 (6.84, 35.86)

Age group

≤30 years 34/434 (8) 1

>30 years 71/313 (23) 2.23 (1.47, 3.38)

Gender

Female 38/352 (11) 1

Male 67/395 (17) 1.64 (1.08, 2.50)

Smoker2

No 84/593 (14) 1

Yes 20/151 (13) 0.64 (0.39, 1.07)

392

Weight loss3

No 11/165 (7) 1

Yes 94/579 (16) 3.51 (1.87, 6.59)

Chest x-ray extent4

Normal/minimal 19/111 (17) 1

Moderate/far advanced 86/628 (14) 0.81 (0.49, 1.34)

1One missed HIV status; 4eight missed extent variable; 2three missed history of ever smoked; 3four missed history of weight loss; low BMI <18.5 kg/m2.

393

Table 13:9 Relative hazards [HR, 95% confidence intervals (CIs)] for death among

HIV positive tuberculosis patients compared with HIV negative patients stratified according to BMI categories

Stratified models

Deaths/N Normal BMI Low BMI Characteristics (%) (n=437) (n=310)

HR (95% CI) HR (95% CI)

HIV status

Negative 6/345 (2) 1 1

Positive 99/401 (25) 42.86 (5.90, 311.47) 10.65 (4.20, 27.02)

Age group

≤30 years 34/434 (8) 1 1

>30 years 71/313 (23) 3.10 (1.63, 5.91) 1.68 (0.97, 2.92)

Gender

Female 38/352 (11) 1 1

Male 67/395 (17) 1.51 (0.83, 2.75) 1.59 (0.87, 2.92)

Smoker2

394

No 84/593 (14) 1 1

Yes 20/151 (13) 0.57 (0.23, 1.40) 0.60 (0.32, 1.12)

Weight loss3

No 11/165 (7) 1 1

Yes 94/579 (16) 3.75 (1.48, 9.54) 3.37 (1.43, 7.97)

Chest x-ray extent4

Normal/minimal 19/111 (17) 1 1

Moderate/far advanced 86/628 (14) 0.98 (0.47, 2.03) 0.63 (0.32, 1.25)

1One missed HIV status; 4eight missed extent variable; 2three missed history of ever smoked; 3four missed history of weight loss; low BMI <18.5 kg/m2.

395

Table 13:10 Multivariable Relative hazards [HR, 95% confidence intervals (CIs)] for death among tuberculosis patients with normal compared with patients having low fat-free mass index (FFMI)

Deaths/N Relative 95% Confidence

Characteristic (%) hazard Interval

Fat-free mass index (kg/m2)

Normal 14/208 (7) 1 -

Low 16/103 (16) - -

Sex

Female 12/147 (8) 1 -

Male 18/164 (11) 3.34 1.23, 9.08

Age (years)

≤30 10/194 (5) 1 -

>30 20/117 (17) 2.24 0.84, 5.95

HIV-serostatus1

Negative 1/172 (0.6) 1 -

Positive 29/138 (21) - -

Hemoglobin

396

>10 mg/dl 14/224 (6) 1 -

≤10 mg/dl 15/74 (20) 1.34 0.52, 3.49

Reduced FFMI*sex 0.13 0.03, 0.56

Reduced FFMI*age 1.61 0.35, 7.34

Reduced FFMI*HIV - -

Reduced FFMI*hemoglobin 2.00 0.49, 8.12

-2LL reduced model - -2LL full model: 332.269 – 321.399 = 10.870; df = 4, p=0.028. 1One missed HIV status; 215 missed hemoglobin measurement due to lack of blood; FFMI low (<16.7 kg/m2 for men and <14.6 kg/m2 for women, normal (≥ 16.7kg/m2 for men, ≥14.6 kg/m2 for women).

397

Table 13:11 Relative hazards [HR, 95% confidence intervals (CIs)] for death among tuberculosis patients with normal compared with low fat-free mass index (FFMI)

Overall model Characteristics Deaths/N (%)

HR (95% CI)

FFMI in kg/m2

Normal (≥18.5) 14/208 (7) 1

Low (<18.5) 16/103 (16) 1.50 (0.73, 3.08)

Gender

Female 12/147 (8) 1

Male 18/164 (11) 1.55 (0.72, 3.37)

Age (years)

≤30 10/194 (5) 1

>30 20/117 (17) 3.57 (1.73, 7.35)

Hemoglobin

2 (mg/dl)

>10 14/224 (6) 1

398

≤10 15/74 (20) 2.85 (1.45, 5.61)

1One missed HIV status; 215 missed hemoglobin measurement due to lack of blood; FFMI low (<16.7 kg/m2 for men and <14.6 kg/m2 for women, normal (≥ 16.7kg/m2 for men, ≥14.6 kg/m2 for women).

399

Table 13:12 Relative hazards [HR, 95% confidence intervals (CIs)] for death among tuberculosis patients with normal compared with low fat-free mass index (FFMI) stratified according to HIV status

Stratified models

HIV negative HIV positive Characteristics Deaths/N (%) (n=186) (n=153)

HR (95% CI) HR (95% CI)

FFMI in kg/m2

Normal (≥18.5) 14/208 (7) 1 1

Low (<18.5) 16/103 (16) - 1.50 (0.72, 3.12)

Gender

Female 12/147 (8) 1 1

Male 18/164 (11) - 1.53 (0.69, 3.39)

Age (years)

≤30 10/194 (5) 1 1

>30 20/117 (17) - 2.21 (1.06, 4.60)

Hemoglobin

2 (mg/dl)

400

>10 14/224 (6) 1 1

≤10 15/74 (20) - 1.65 (0.82, 3.30)

1One missed HIV status; 215 missed hemoglobin measurement due to lack of blood; FFMI low (<16.7 kg/m2 for men and <14.6 kg/m2 for women, normal (≥ 16.7kg/m2 for men, ≥14.6 kg/m2 for women).

401

Table 13:13 Relative hazards [HR, 95% confidence intervals (CIs)] for death among tuberculosis patients with normal compared with low fat-free mass index (FFMI)

Overall model Characteristics Deaths/N (%)

HR (95% CI)

FFMI in kg/m2

Normal (≥18.5) 14/208 (7) 1

Low (<18.5) 16/103 (16) 1.55 (0.76, 3.15)

Gender

Female 12/147 (8) 1

Male 18/164 (11) 1.92 (0.90, 4.11)

HIV-serostatus1

Negative 1/172 (0.6) 1

Positive 29/138 (21) 41.68 (5.67, 306.41)

Hemoglobin

2 (mg/dl)

>10 14/224 (6) 1

402

≤10 15/74 (20) 1.79 (0.90, 3.55)

1One missed HIV status; 215 missed hemoglobin measurement due to lack of blood; FFMI low (<16.7 kg/m2 for men and <14.6 kg/m2 for women, normal (≥ 16.7kg/m2 for men, ≥14.6 kg/m2 for women).

403

Table 13:14 Relative hazards [HR, 95% confidence intervals (CIs)] for death among

tuberculosis patients with normal compared with low fat-free mass index (FFMI) stratified according to age group

Stratified models

Young ≤30 years Old >30 years Characteristics Deaths/N (%) (n=212) (n=128)

HR (95% CI) HR (95% CI)

FFMI in kg/m2

Normal (≥18.5) 14/208 (7) 1 1

Low (<18.5) 16/103 (16) 1.89 (0.55, 6.52) 1.52 (0.63, 3.69)

Gender

Female 12/147 (8) 1 1

Male 18/164 (11) 1.62 (0.45, 5.80) 1.56 (0.58, 4.19)

HIV-serostatus1

Negative 1/172 (0.6) 1 1

Positive 29/138 (21) - 16.03 (2.14, 119.84)

Hemoglobin

2 (mg/dl)

404

>10 14/224 (6) 1 1

≤10 15/74 (20) 1.43 (0.42, 4.91) 2.02 (0.89, 4.58)

1One missed HIV status; 215 missed hemoglobin measurement due to lack of blood; FFMI low (<16.7 kg/m2 for men and <14.6 kg/m2 for women, normal (≥ 16.7kg/m2 for men, ≥14.6 kg/m2 for women).

405

Table 13:15 Relative hazards [HR, 95% confidence intervals (CIs)] for death among

HIV positive tuberculosis patients compared with HIV negative patients

Overall model Characteristics Deaths/N (%)

HR (95% CI)

HIV status

Negative 1/172 (0.6) 1

Positive 29/138 (21) 34.37 (4.65, 254.20)

Age (years)

≤30 10/194 (5) 1

>30 20/117 (17) 2.33 (1.12, 4.83)

Gender

Female 12/147 (8) 1

Male 18/164 (11) 1.89 (0.91, 3.91)

Hemoglobin

2 (mg/dl)

>10 14/224 (6) 1

406

≤10 15/74 (20) 1.86 (0.94, 3.67)

1One missed HIV status; 215 missed hemoglobin measurement due to lack of blood; FFMI low (<16.7 kg/m2 for men and <14.6 kg/m2 for women, normal (≥ 16.7kg/m2 for men, ≥14.6 kg/m2 for women).

407

Table 13:16 Relative hazards [HR, 95% confidence intervals (CIs)] for death among

HIV positive tuberculosis patients compared with HIV negative patients stratified according to fat-free mass index (FFMI) categories

Stratified models

Normal FFMI Characteristics Deaths/N (%) Low FFMI (n=103) (n=208)

HR (95% CI) HR (95% CI)

HIV status

Negative 1/172 (0.6) 1 1

Positive 29/138 (21) - 14.80 (1.93, 113.58)

Age (years)

≤30 10/194 (5) 1 1

>30 20/117 (17) 2.04 (0.68, 6.10) 3.50 (1.10, 11.11)

Gender

Female 12/147 (8) 1 1

Male 18/164 (11) 2.28 (0.77, 6.75) 0.43 (0.14, 1.31)

Hemoglobin

2 (mg/dl)

408

>10 14/224 (6) 1 1

≤10 15/74 (20) 1.50 (0.51, 4.42) 2.64 (0.95, 7.39)

1One missed HIV status; 215 missed hemoglobin measurement due to lack of blood; FFMI low (<16.7 kg/m2 for men and <14.6 kg/m2 for women, normal (≥ 16.7kg/m2 for men, ≥14.6 kg/m2 for women).

409

Figure 13:1 Survival distribution among women with low baseline body mass index

(BMI) (<18.5 kg/m2) compared to women with normal BMI in Uganda

410

Figure 13:2 Survival distribution among men with low (<16.7 kg/m2 for men, <14.6 kg/m2 for women) baseline fat-free mass index (FFMI) compared to men with normal FFMI in Uganda

411

Figure 13:3 Survival distribution among Men compared to Women tuberculosis patients in urban Uganda

Data truncated at 1280 days of follow-up

412

Figure 13:4 Survival distribution among Men compared with Women among HIV negative tuberculosis patients in Uganda

413

CHAPTER 14

EXTRA TABLES AND FIGURES FOR CHAPTER TEN

Table 14:1 Spearman raw correlation matrix for fat-free mass, fat mass, and body mass index over different visit time points

414

Visit

(Month)

Fat-free mass

0 Fat-free mass 1

3 Fat-free mass 0.73 1

12 Fat-free mass 0.69 0.78 1

24 Fat-free mass 0.68 0.72 0.78

Fat mass

0 Fat mass 1

3 Fat mass 0.88 1

12 Fat mass 0.85 0.87 1

24 Fat mass 0.81 0.85 0.90

Body mass index

0 Body mass index 1

2 Body mass index 0.83 1

3 Body mass index 0.88 0.91 1

5 Body mass index 0.82 0.87 0.90 1

415

6 Body mass index 0.68 0.77 077 0.77 1

12 Body mass index 0.82 0.83 0.86 0.85 0.74 1

24 Body mass index 0.75 0.80 0.80 0.83 0.71 0.89

Table 14:2 Regression estimates for effects of baseline wasting, lag FFMI measure, and baseline characteristics on probability of missing

416

Covariate Estimate Standard error 95% CI

Proc GENMOD FFMI

Intercept -1.04 1.79 - 4.56, 2.48

Baseline wasting -0.34 0.41 -1.25, 0.47

Baseline FFMI -0.03 0.12 -0.26, 0.20

Lag FFMI -0.11 0.10 -1.07, 0.29

Sex 0.28 0.28 -0.27, 0.82

Age 0.15 0.20 -0.25, 0.54

HIV 0.42 0.20 2.10, 0.04

Smoker 0.29 0.26 -0.23, 0.81

Hemoglobin -0.42 0.23 -0.87, 0.02

History of weight loss 0.09 0.25 -0.40, 0.58

Chest x-ray extent -0.11 0.27 -0.63, 0.41

Time*baseline wasting 0.02 0.02 -0.02, 0.06

Proc MIXED FFMI

Intercept 15.65 0.50 14.67, 16.63

Probability missing -0.06 0.12 -0.30, 0.18

417

Baseline wasting -1.68 0.14 -1.95, -1.41

Lag FFMI 0.04 0.03 -0.02, 0.09

Sex 1.74 0.13 1.48, 2.00

Age 0.10 0.12 -0.13, 0.34

HIV 0.06 0.12 -0.18, 0.29

Smoker -0.14 0.15 -0.44, 0.17

Hemoglobin -0.34 0.14 -0.60, -0.07

History of weight loss -0.03 0.15 -0.33, 0.28

Chest x-ray extent -0.10 0.16 -0.40, 0.21

Time*missing -0.01 0.01 -0.02, 0.01

Random intercept (b0i) variance 0.458 0.098 <0.001*

Random slope (b1i) variance 0.007 0.004 0.133*

Covariance slopes (b0i, b1i) 0.0004 0.0003 0.097*

Autocorrelation; exponential 0.8610 0 -

Residual (measurement error) 0.567 0.053 <0.001*

Table 14:3 Regression estimates for effects of baseline wasting, lag FMI measure, and baseline characteristics on probability of missing

418

Covariate Estimate Standard error 95% CI

Proc GENMOD FMI

Intercept -3.84 0.63 - 5.08, -2.59

Baseline wasting 0.45 0.40 -0.34, 1.24

Baseline FMI -0.10 0.07 -0.03, 0.24

Lag FMI -0.04 0.05 -0.14, 0.06

Sex 0.25 0.31 -0.35, 0.84

Age 0.16 0.20 -0.23, 0.55

HIV 0.44 0.20 0.05, 0.83

Smoker 0.32 0.27 -0.20, 0.84

Hemoglobin -0.29 0.24 -0.76, 0.17

History of weight loss 0.17 0.26 -0.34, 0.67

Chest x-ray extent -0.01 0.27 -0.53, 0.51

Time*baseline wasting -0.02 0.02 -0.06, 0.03

Proc MIXED FMI

Intercept 6.37 0.36 5.66, 7.08

Probability missing 0.31 0.20 -0.08, 0.70

419

Baseline wasting -2.18 0.20 -2.57, -1.79

Lag FMI 0.03 0.02 -0.01, 0.07

Sex -2.91 0.21 -3.32, -2.49

Age -0.02 0.20 -0.42, 0.38

HIV -0.14 0.20 -0.54, 0.25

Smoker 0.24 0.26 -0.26, 0.75

Hemoglobin -0.44 0.24 -0.90, 0.03

History of weight loss -0.52 0.26 -1.03, -0.02

Chest x-ray extent -0.45 0.26 -0.96, 0.06

Time*missing 0.005 0.01 -0.02, 0.03

Random intercept (b0i) 1.721 0.207 <0.001* variance

Random slope (b1i) 0.004 0.009 0.642* variance

Covariance slopes (b0i, b1i) 0.003 0.0006 <0.001*

Autocorrelation; 0.9997 3.383E8 0.500* exponential

0.684 0.061 <0.001* Residual (measurement

420

error)

*p-values

421

Table 14:4 Assessing contributions of polynomials, random intercepts, and random slopes in mixed models

Fat-free mass

Parms Dif - p-

Model AIC -2RLL (df+1) 2RLL df value

Linear model 2768.2 2762.2 3 - - -

Quadratic model 2751.4 2745.4 3 16.8

Cubic model 2730.3 2724.3 3 37.9

Fixed intercepts 2691.6 2673.6 3 - - -

Random 0 0 NA intercepts 2691.6 2673.6 3

Random

Intercepts and 27 2 <0.001 slopes 2668.6 2646.6 5

Fat mass

Linear model 3197.6 3191.6 3 - - -

Quadratic model 3182.0 3176.0 3 15.6 0 NA

Cubic model 3190.5 3184.5 3 7.1 0 NA

422

Fixed intercepts 3155.6 3137.6 3 - - -

Random 0 intercepts 3155.6 3137.6 3

Random

intercepts and 52.2 2 <0.001 slopes 3107.4 3085.4 5

Body mass index

Linear model 15923.8 15917.8 3 - - -

Quadratic model 15768.2 15762.2 3 155.6 0 NA

Cubic model 15688.8 15682.8 3 235.0 0 NA

Fixed intercepts 15650.0 15632.0 3 - - -

Random 5.6 1 intercepts 15646.4 15626.4 4

Random intercepts and 116.5 3 <0.001 slopes 15539.5 15515.5 6

AIC = Akaike Information Criteria, -2LL = -2 log likelihood, df = degrees of freedom, NA = not applicable.

423

Table 14:5 Assessing Piecewise and polynomial models; random intercepts, and random slopes in mixed models

Fat-free mass

Parms Dif - p-

Model AIC -2RLL (df+1) 2RLL df value

Linear model 2779.4 2767.4 6 - - -

Quadratic model 2755.6 2737.6 9 29.8 3 <0.001

2 Spline model 2714.8 2686.8 15 NA NA NA

Random intercepts 2734.7 2722.7 6 - - -

Random early slope 2715.8 2699.8 8 22.9 2 <0.001

Random early & 3 months slopes 2709.5 2687.5 11 35.2 5 <0.001

Random early, 3, &

12 months slopes 2714.8 2686.8 15 35.9 9 <0.001

Fat mass

Linear model 3210.4 3202.4 6 - - -

Quadratic model 3276.4 3262.4 9 60 3 <0.001

424

2 Spline model 3274.8 3252.8 15 NA NA NA

Random intercepts 3357.2 3361.2 6 - - -

Random early slope 3273.2 3265.2 8 96 2 <0.001

Random early & 3 months slopes 3276.4 3264.4 11 96.8 4 <0.001

Random early, 3, &

12 months slopes 3274.8 3252.8 15 108.4 9 <0.001

Body mass index

Linear model 16307.3 16295.3 6 - - -

Quadratic model 16003.5 15983.5 9 311.8 3 <0.001

2 Spline model 15887.2 15859.2 15 NA NA NA

Random intercepts 16079.2 16067.2 6 - - -

Random early slope 15911.8 15895.8 8 171.4 2 <0.001

Random early & 3 months slopes 15894.2 15872.2 11 195.0 5 <0.001

Random early, 3, &

12 months slopes 15887.2 15859.2 15 208.0 8 <0.001

AIC = Akaike Information Criteria, -2LL = -2 log likelihood, df = degrees of freedom, NA = not applicable.

425

Figure 14:1 Individual profiles for fat-free mass over time

426

Figure 14:2 Individual profiles for fat mass index over time

427

Figure 14:3 Individual profiles for BMI over time

428

Figure 14:4 Box plots for fat-free mass index over time

429

Figure 14:5 Box plots for fat mass index over time

430

Figure 14:6 Box plots for BMI over time

431

Figure 14:7 Mean profile for FFMI among patients with compared to patients without baseline wasting

432

Figure 14:8 Mean profile for FFMI among men compared to mean profile among women

433

Figure 14:9 Mean profile for FFMI among HIV negative compared to mean profile among HIV positive patients

434

Figure 14:10 Mean profile for FMI among patients with compared to patients without wasting

435

Figure 14:11 Mean profile for FMI among men compared to mean profile among women

436

Figure 14:12 Mean profile for FMI among HIV negative compared to mean profile among HIV positive patients

437

Figure 14:13 Mean profile for BMI among patients with compared to patients without wasting

438

Figure 14:14 Mean profiles for BMI among men compared to mean profile among women

439

Figure 14:15 Mean profile for BMI among HIV negative compared to mean profile among HIV positive patients

440

BIBLIOGRAPHY

CHAPTER ONE

Chandra, R. K. 1991. 1990 McCollum Award lecture. Nutrition and immunity: lessons

from the past and new insights into the future. Am J Clin Nutr 53 (5):1087-101.

Fernandes G, Jolly C.A, and Lawrence R.A. 2006. Nutrition and The Immune System. In

Modern Nutrition in Health and Disease, edited by Shills M.E, Shike M, Ross

C.A, Caballero B and Cousins R.J. Philadelphia PA: Lippincott Williams &

Wilkins.

Gershwin M.E, Beach R.S, and Hurley L.S. 1985. Nutrition and Immunity. Olando FL:

Academic Press.

Harries, A. D., W. A. Nkhoma, P. J. Thompson, D. S. Nyangulu, and J. J. Wirima. 1988.

Nutritional status in Malawian patients with pulmonary tuberculosis and response

to chemotherapy. Eur J Clin Nutr 42 (5):445-50.

Kennedy, N., A. Ramsay, L. Uiso, J. Gutmann, F. I. Ngowi, and S. H. Gillespie. 1996.

Nutritional status and weight gain in patients with pulmonary tuberculosis in

Tanzania. Trans R Soc Trop Med Hyg 90 (2):162-6.

Kyle, U. G., I. Bosaeus, A. D. De Lorenzo, P. Deurenberg, M. Elia, J. Manuel Gomez, B.

Lilienthal Heitmann, L. Kent-Smith, J. C. Melchior, M. Pirlich, H. Scharfetter, M.

W. J. Schols A, and C. Pichard. 2004. Bioelectrical impedance analysis-part II:

utilization in clinical practice. Clin Nutr 23 (6):1430-53.

441

Kyle, U. G., L. Genton, and C. Pichard. 2002. Body composition: what's new? Curr Opin

Clin Nutr Metab Care 5 (4):427-33.

Kyle, U. G., A. Piccoli, and C. Pichard. 2003. Body composition measurements:

interpretation finally made easy for clinical use. Curr Opin Clin Nutr Metab Care

6 (4):387-93.

Lawn, S. D., and G. Churchyard. 2009. Epidemiology of HIV-associated tuberculosis.

Curr Opin HIV AIDS 4 (4):325-33.

Lucas, S. B., K. M. De Cock, A. Hounnou, C. Peacock, M. Diomande, M. Honde, A.

Beaumel, L. Kestens, and A. Kadio. 1994. Contribution of tuberculosis to slim

disease in Africa. Bmj 308 (6943):1531-3.

Lucas, S. B., A. Hounnou, C. Peacock, A. Beaumel, G. Djomand, J. M. N'Gbichi, K.

Yeboue, M. Honde, M. Diomande, C. Giordano, and et al. 1993. The mortality

and pathology of HIV infection in a west African city. Aids 7 (12):1569-79.

Mostert, R., A. Goris, C. Weling-Scheepers, E. F. Wouters, and A. M. Schols. 2000.

Tissue depletion and health related quality of life in patients with chronic

obstructive pulmonary disease. Respir Med 94 (9):859-67.

Mugusi, F. M., S. Mehta, E. Villamor, W. Urassa, E. Saathoff, R. J. Bosch, and W. W.

Fawzi. 2009. Factors associated with mortality in HIV-infected and uninfected

patients with pulmonary tuberculosis. BMC Public Health 9:409.

Mupere, E., S. Zalwango, A. Chiunda, A. Okwera, R. Mugerwa, and C. Whalen. 2010.

Body composition among HIV-seropositive and HIV-seronegative adult patients

with pulmonary tuberculosis in Uganda. Ann Epidemiol 20 (3):210-6.

442

Niyongabo, T., D. Henzel, M. Idi, S. Nimubona, E. Gikoro, J. C. Melchior, S. Matheron,

G. Kamanfu, B. Samb, B. Messing, J. Begue, P. Aubry, and B. Larouze. 1999.

Tuberculosis, human immunodeficiency virus infection, and malnutrition in

Burundi. Nutrition 15 (4):289-93.

Nunn, P., R. Brindle, L. Carpenter, J. Odhiambo, K. Wasunna, R. Newnham, W. Githui,

S. Gathua, M. Omwega, and K. McAdam. 1992. Cohort study of human

immunodeficiency virus infection in patients with tuberculosis in Nairobi, Kenya.

Analysis of early (6-month) mortality. Am Rev Respir Dis 146 (4):849-54.

Paton, N. I., and Y. M. Ng. 2006. Body composition studies in patients with wasting

associated with tuberculosis. Nutrition 22 (3):245-51.

Ramakrishnan, C. V., K. Rajendran, P. G. Jacob, W. Fox, and S. Radhakrishna. 1961.

The role of diet in the treatment of pulmonary tuberculosis. An evaluation in a

controlled chemotherapy study in home and sanatorium patients in South India.

Bull World Health Organ 25:339-59.

Scalcini, M., R. Occenac, J. Manfreda, and R. Long. 1991. Pulmonary tuberculosis,

human immunodeficiency virus type-1 and malnutrition. Bull Int Union Tuberc

Lung Dis 66 (1):37-41.

Suttmann, U., J. Ockenga, O. Selberg, L. Hoogestraat, H. Deicher, and M. J. Muller.

1995. Incidence and prognostic value of malnutrition and wasting in human

immunodeficiency virus-infected outpatients. J Acquir Immune Defic Syndr Hum

Retrovirol 8 (3):239-46.

443

van Lettow, M., W. W. Fawzi, and R. D. Semba. 2003. Triple trouble: the role of

malnutrition in tuberculosis and human immunodeficiency virus co-infection.

Nutr Rev 61 (3):81-90.

Vitoria, M., R. Granich, C. F. Gilks, C. Gunneberg, M. Hosseini, W. Were, M.

Raviglione, and K. M. De Cock. 2009. The global fight against HIV/AIDS,

tuberculosis, and malaria: current status and future perspectives. Am J Clin Pathol

131 (6):844-8.

Wagner, G. J., S. J. Ferrando, and J. G. Rabkin. 2000. Psychological and physical health

correlates of body cell mass depletion among HIV+ men. J Psychosom Res 49

(1):55-7.

World Health Organization. Global tuberculosis control. Epidemiology, strategy,

financing. WHO/HTM/TB/2009.411. 2009. Geneva: World Health Organization

Zachariah, R., M. P. Spielmann, A. D. Harries, and F. M. Salaniponi. 2002. Moderate to

severe malnutrition in patients with tuberculosis is a risk factor associated with

early death. Trans R Soc Trop Med Hyg 96 (3):291-4.

444

CHAPTER TWO

Arkhipova O.P. 1975. Impact of tuberculosis infection and antibacterial preparations on

thiamin metabolism. Voprosy Meditsinskoi Khimii 21:449 - 560.

Baum, M. K., E. Mantero-Atienza, G. Shor-Posner, M. A. Fletcher, R. Morgan, C.

Eisdorfer, H. E. Sauberlich, P. E. Cornwell, and R. S. Beach. 1991. Association of

vitamin B6 status with parameters of immune function in early HIV-1 infection. J

Acquir Immune Defic Syndr 4 (11):1122-32.

Beach, R. S., E. Mantero-Atienza, G. Shor-Posner, J. J. Javier, J. Szapocznik, R. Morgan,

H. E. Sauberlich, P. E. Cornwell, C. Eisdorfer, and M. K. Baum. 1992. Specific

nutrient abnormalities in asymptomatic HIV-1 infection. Aids 6 (7):701-8.

Bhaskaram, P. 1992. The vicious cycle of malnutrition-infection with special reference to

diarrhea, measles and tuberculosis. Indian Pediatr 29 (6):805-14.

Bhuyan, U. N., and V. Ramalingaswami. 1973. Immune responses of the protein-

deficient guinea pig to BCG vaccination. Am J Pathol 72 (3):489-502.

Bogden, J. D., H. Baker, O. Frank, G. Perez, F. Kemp, K. Bruening, and D. Louria. 1990.

Micronutrient status and human immunodeficiency virus (HIV) infection. Ann N

Y Acad Sci 587:189-95.

Cegielski, J. P., and D. N. McMurray. 2004. The relationship between malnutrition and

tuberculosis: evidence from studies in humans and experimental animals. Int J

Tuberc Lung Dis 8 (3):286-98.

445

Chan J, Tanaka K.E, Mannion C, Carroll D, Tsang M.S, Xing Y, Lowenstein C, and

Bloom B.R. 1997. Effects of protein calorie malnutrition on mice infected with

BCG. J Nutr Immunol.

Chandra, R. K. 1991. 1990 McCollum Award lecture. Nutrition and immunity: lessons

from the past and new insights into the future. Am J Clin Nutr 53 (5):1087-101.

Chandra, R. K., and S. Kumari. 1994. Nutrition and immunity: an overview. J Nutr 124

(8 Suppl):1433S-1435S.

Comstock, G. W., and C. E. Palmer. 1966. Long-term results of BCG vaccination in the

southern United States. Am Rev Respir Dis 93 (2):171-83.

Ehrenpreis, E. D., S. J. Carlson, H. L. Boorstein, and R. M. Craig. 1994. Malabsorption

and deficiency of vitamin B12 in HIV-infected patients with chronic diarrhea. Dig

Dis Sci 39 (10):2159-62.

Evans, D. I., and B. Attock. 1971. Folate deficiency in pulmonary tuberculosis:

relationship to treatment and to serum vitamin A and beta-carotene. Tubercle 52

(4):288-94.

Harries, A. D., W. A. Nkhoma, P. J. Thompson, D. S. Nyangulu, and J. J. Wirima. 1988.

Nutritional status in Malawian patients with pulmonary tuberculosis and response

to chemotherapy. Eur J Clin Nutr 42 (5):445-50.

Harriman, G. R., P. D. Smith, M. K. Horne, C. H. Fox, S. Koenig, E. E. Lack, H. C.

Lane, and A. S. Fauci. 1989. Vitamin B12 malabsorption in patients with acquired

immunodeficiency syndrome. Arch Intern Med 149 (9):2039-41.

446

Karyadi, E., W. Schultink, R. H. Nelwan, R. Gross, Z. Amin, W. M. Dolmans, J. W. van

der Meer, J. G. Hautvast, and C. E. West. 2000. Poor micronutrient status of

active pulmonary tuberculosis patients in Indonesia. J Nutr 130 (12):2953-8.

Karyadi, E., C. E. West, W. Schultink, R. H. Nelwan, R. Gross, Z. Amin, W. M.

Dolmans, H. Schlebusch, and J. W. van der Meer. 2002. A double-blind, placebo-

controlled study of vitamin A and zinc supplementation in persons with

tuberculosis in Indonesia: effects on clinical response and nutritional status. Am J

Clin Nutr 75 (4):720-7.

Keusch, G. T. 1990. Micronutrients and susceptibility to infection. Ann N Y Acad Sci

587:181-8.

Line, D. H., B. Seitanidis, J. O. Morgan, and A. V. Hoffbrand. 1971. The effects of

chemotherapy on iron, folate, and vitamin B 12 metabolism in tuberculosis. Q J

Med 40 (159):331-40.

Lucas, S. B., K. M. De Cock, A. Hounnou, C. Peacock, M. Diomande, M. Honde, A.

Beaumel, L. Kestens, and A. Kadio. 1994. Contribution of tuberculosis to slim

disease in Africa. Bmj 308 (6943):1531-3.

Macallan, D. C. 1999. Malnutrition in tuberculosis. Diagn Microbiol Infect Dis 34

(2):153-7.

Macallan, D. C., M. A. McNurlan, A. V. Kurpad, G. de Souza, P. S. Shetty, A. G. Calder,

and G. E. Griffin. 1998. Whole body protein metabolism in human pulmonary

tuberculosis and undernutrition: evidence for anabolic block in tuberculosis. Clin

Sci (Lond) 94 (3):321-31.

447

Macallan, D. C., M. A. McNurlan, E. Milne, A. G. Calder, P. J. Garlick, and G. E.

Griffin. 1995. Whole-body protein turnover from leucine kinetics and the

response to nutrition in human immunodeficiency virus infection. Am J Clin Nutr

61 (4):818-26.

Macallan, D. C., C. Noble, C. Baldwin, S. A. Jebb, A. M. Prentice, W. A. Coward, M. B.

Sawyer, T. J. McManus, and G. E. Griffin. 1995. Energy expenditure and wasting

in human immunodeficiency virus infection. N Engl J Med 333 (2):83-8.

Madebo, T., G. Nysaeter, and B. Lindtjorn. 1997. HIV infection and malnutrition change

the clinical and radiological features of pulmonary tuberculosis. Scand J Infect

Dis 29 (4):355-9.

Markkanen, T., A. Levanto, V. Sallinen, and S. Virtanen. 1967. Folic acid and vitamin

B12 in tuberculosis. Scand J Haematol 4 (4):283-91.

McMurray, D. N., R. A. Bartow, C. L. Mintzer, and E. Hernandez-Frontera. 1990.

Micronutrient status and immune function in tuberculosis. Ann N Y Acad Sci

587:59-69.

Miansikov V.G. 1969. Some indices of vitamion B6 metabolism in patients with

pulmonary tuberculosis in elderly and old persons. Vrachebnoe Delo 3:77 - 79.

Miller, L. G., S. M. Asch, E. I. Yu, L. Knowles, L. Gelberg, and P. Davidson. 2000. A

population-based survey of tuberculosis symptoms: how atypical are atypical

presentations? Clin Infect Dis 30 (2):293-9.

Mupere, E., S. Zalwango, A. Chiunda, A. Okwera, R. Mugerwa, and C. Whalen. 2010.

Body composition among HIV-seropositive and HIV-seronegative adult patients

with pulmonary tuberculosis in Uganda. Ann Epidemiol 20 (3):210-6.

448

Onwubalili, J. K. 1988. Malnutrition among tuberculosis patients in Harrow, England.

Eur J Clin Nutr 42 (4):363-6.

Panasyuk A.V, Penenko O.R, Kuzmenko I.V, and et al. 1991. Vitamin E and its structural

analogs in tuberculosis. Urkr Biokhim Zh 63:83 - 88.

Paton, N. I., L. R. Castello-Branco, G. Jennings, M. B. Ortigao-de-Sampaio, M. Elia, S.

Costa, and G. E. Griffin. 1999. Impact of tuberculosis on the body composition of

HIV-infected men in Brazil. J Acquir Immune Defic Syndr Hum Retrovirol 20

(3):265-71.

Paton, N. I., Y. K. Chua, A. Earnest, and C. B. Chee. 2004. Randomized controlled trial

of nutritional supplementation in patients with newly diagnosed tuberculosis and

wasting. Am J Clin Nutr 80 (2):460-5.

Paton, N. I., and Y. M. Ng. 2006. Body composition studies in patients with wasting

associated with tuberculosis. Nutrition 22 (3):245-51.

Paton, N. I., Y. M. Ng, C. B. Chee, C. Persaud, and A. A. Jackson. 2003. Effects of

tuberculosis and HIV infection on whole-body protein metabolism during feeding,

measured by the [15N]glycine method. Am J Clin Nutr 78 (2):319-25.

Rao K.N, and Gopalan C. 1966. The role of nutritional factors in tuberculosis. Indian J

Tuberculosis 13:102 - 106.

Rook, G. A., and R. Hernandez-Pando. 1996. The pathogenesis of tuberculosis. Annu Rev

Microbiol 50:259-84.

Safarian, M. D., K. G. Karagezian, E. T. Karapetian, and N. A. Avanesian. 1990. [The

efficacy of antioxidant therapy in patients with tuberculosis of the lungs and the

correction of lipid peroxidation processes]. Probl Tuberk (5):40-4.

449

Sarraf, P., R. C. Frederich, E. M. Turner, G. Ma, N. T. Jaskowiak, D. J. Rivet, 3rd, J. S.

Flier, B. B. Lowell, D. L. Fraker, and H. R. Alexander. 1997. Multiple cytokines

and acute inflammation raise mouse leptin levels: potential role in inflammatory

anorexia. J Exp Med 185 (1):171-5.

Scalcini, M., R. Occenac, J. Manfreda, and R. Long. 1991. Pulmonary tuberculosis,

human immunodeficiency virus type-1 and malnutrition. Bull Int Union Tuberc

Lung Dis 66 (1):37-41.

Schwenk A, and Macallan D.C. 2000. Tuberculosis, malnutrition and wasting. Curr Opin

Clin Nutr Metab Care 3:285 - 91.

Schwenk, A., L. Hodgson, A. Wright, L. C. Ward, C. F. Rayner, S. Grubnic, G. E.

Griffin, and D. C. Macallan. 2004. Nutrient partitioning during treatment of

tuberculosis: gain in body fat mass but not in protein mass. Am J Clin Nutr 79

(6):1006-12.

Semba, R. D., and A. M. Tang. 1999. Micronutrients and the pathogenesis of human

immunodeficiency virus infection. Br J Nutr 81 (3):181-9.

Smurova, T. F., and D. I. Prokop'ev. 1969. [Vitamin A and carotene levels in the blood of

pati-ents with pulmonary tuberculosis and diabetes mellitus]. Probl Tuberk 47

(11):50-5.

Tverdal, A. 1986. Body mass index and incidence of tuberculosis. Eur J Respir Dis 69

(5):355-62.

Ullrich, R., T. Schneider, W. Heise, W. Schmidt, R. Averdunk, E. O. Riecken, and M.

Zeitz. 1994. Serum carotene deficiency in HIV-infected patients. Berlin

Diarrhoea/Wasting Syndrome Study Group. Aids 8 (5):661-5.

450

van Lettow, M., W. W. Fawzi, and R. D. Semba. 2003. Triple trouble: the role of

malnutrition in tuberculosis and human immunodeficiency virus co-infection.

Nutr Rev 61 (3):81-90.

Verbon, A., N. Juffermans, S. J. Van Deventer, P. Speelman, H. Van Deutekom, and T.

Van Der Poll. 1999. Serum concentrations of cytokines in patients with active

tuberculosis (TB) and after treatment. Clin Exp Immunol 115 (1):110-3.

Volosevich, G. V. 1982. [Blood T-lymphocyte functional activity in pulmonary

tuberculosis patients undergoing chemotherapeutic agent and vitamin treatment].

Probl Tuberk (3):47-50.

Whalen C, and Semba R.D. 2001. Tuberculosis. In Nutrition and Health in Developing

Countries, edited by Semba R.D and Bloem M.W. Totowa NJ, Human Press.

Zacharia R, M. P Spielmann, Harries A.D, and Salaniponi F.M. 2002. Malnutrition in

tuberculosis patients on admission and weight-gain in relation to HIV status in

Thyolo district of Malawi. Malawi Med J 13:12 - 13.

Zachariah, R., M. P. Spielmann, A. D. Harries, and F. M. Salaniponi. 2002. Moderate to

severe malnutrition in patients with tuberculosis is a risk factor associated with

early death. Trans R Soc Trop Med Hyg 96 (3):291-4.

451

CHAPTER THREE

Guwatudde, D., M. Nakakeeto, E. C. Jones-Lopez, A. Maganda, A. Chiunda, R. D.

Mugerwa, J. J. Ellner, G. Bukenya, and C. C. Whalen. 2003. Tuberculosis in

household contacts of infectious cases in Kampala, Uganda. Am J Epidemiol 158

(9):887-98.

International Union Against Tuberculosis and Lung Disease. Technical guide for sputum

examination for tuberculosis by direct microscopy. 1986. Bull Int Union Tuberc

Lung Dis 61:1 - 16.

Jackson, A. S., M. L. Pollock, J. E. Graves, and M. T. Mahar. 1988. Reliability and

validity of bioelectrical impedance in determining body composition. J Appl

Physiol 64 (2):529-34.

Kotler, D. P., S. Burastero, J. Wang, and R. N. Pierson, Jr. 1996. Prediction of body cell

mass, fat-free mass, and total body water with bioelectrical impedance analysis:

effects of race, sex, and disease. Am J Clin Nutr 64 (3 Suppl):489S-497S.

Kyle, U. G., I. Bosaeus, A. D. De Lorenzo, P. Deurenberg, M. Elia, J. Manuel Gomez, B.

Lilienthal Heitmann, L. Kent-Smith, J. C. Melchior, M. Pirlich, H. Scharfetter, M.

W. J. Schols A, and C. Pichard. 2004. Bioelectrical impedance analysis-part II:

utilization in clinical practice. Clin Nutr 23 (6):1430-53.

Kyle, U. G., L. Genton, and C. Pichard. 2002. Body composition: what's new? Curr Opin

Clin Nutr Metab Care 5 (4):427-33.

452

Kyle, U. G., A. Piccoli, and C. Pichard. 2003. Body composition measurements:

interpretation finally made easy for clinical use. Curr Opin Clin Nutr Metab Care

6 (4):387-93.

Mayanja-Kizza, H., E. Jones-Lopez, A. Okwera, R. S. Wallis, J. J. Ellner, R. D.

Mugerwa, and C. C. Whalen. 2005. Immunoadjuvant prednisolone therapy for

HIV-associated tuberculosis: a phase 2 clinical trial in Uganda. J Infect Dis 191

(6):856-65.

Physical status: the use and interpretation of anthropometry. Report of a WHO Expert

Committee. World Health Organ Tech Rep Ser, 854: 1 - 452. 1995. Geneva.

Schutz, Y., U. U. Kyle, and C. Pichard. 2002. Fat-free mass index and fat mass index

percentiles in Caucasians aged 18-98 y. Int J Obes Relat Metab Disord 26

(7):953-60.

Shah, S., C. Whalen, D. P. Kotler, H. Mayanja, A. Namale, G. Melikian, R. Mugerwa,

and R. D. Semba. 2001. Severity of human immunodeficiency virus infection is

associated with decreased phase angle, fat mass and body cell mass in adults with

pulmonary tuberculosis infection in Uganda. J Nutr 131 (11):2843-7.

Stein, C. M., L. Nshuti, A. B. Chiunda, W. H. Boom, R. C. Elston, R. D. Mugerwa, S. K.

Iyengar, and C. C. Whalen. 2005. Evidence for a major gene influence on tumor

necrosis factor-alpha expression in tuberculosis: path and segregation analysis.

Hum Hered 60 (2):109-18.

Van Lettow, M., J. J. Kumwenda, A. D. Harries, C. C. Whalen, T. E. Taha, N.

Kumwenda, C. Kang'ombe, and R. D. Semba. 2004. Malnutrition and the severity

453

of lung disease in adults with pulmonary tuberculosis in Malawi. Int J Tuberc

Lung Dis 8 (2):211-7.

VanItallie, T. B., M. U. Yang, S. B. Heymsfield, R. C. Funk, and R. A. Boileau. 1990.

Height-normalized indices of the body's fat-free mass and fat mass: potentially

useful indicators of nutritional status. Am J Clin Nutr 52 (6):953-9.

Villamor, E., E. Saathoff, F. Mugusi, R. J. Bosch, W. Urassa, and W. W. Fawzi. 2006.

Wasting and body composition of adults with pulmonary tuberculosis in relation

to HIV-1 coinfection, socioeconomic status, and severity of tuberculosis. Eur J

Clin Nutr 60 (2):163-71.

454

CHAPTER FOUR

Bisai, S., and K. Bose. 2009. Undernutrition in the Kora Mudi tribal population, West

Bengal, India: a comparison of body mass index and mid-upper-arm

circumference. Food Nutr Bull 30 (1):63-7.

Blaak, E. 2001. Gender differences in fat metabolism. Curr Opin Clin Nutr Metab Care 4

(6):499-502.

Bland, J. M., and D. G. Altman. 1986. Statistical methods for assessing agreement

between two methods of clinical measurement. Lancet 1 (8476):307-10.

Briend, A., M. Garenne, B. Maire, O. Fontaine, and K. Dieng. 1989. Nutritional status,

age and survival: the muscle mass hypothesis. Eur J Clin Nutr 43 (10):715-26.

Collins, S. 1996. Using middle upper arm circumference to assess severe adult

malnutrition during famine. Jama 276 (5):391-5.

Dioum, A., A. Gartner, B. Maire, F. Delpeuch, and S. Wade. 2005. Body composition

predicted from skinfolds in African women: a cross-validation study using air-

displacement plethysmography and a black-specific equation. Br J Nutr 93

(6):973-9.

Durnin, J. V., and J. Womersley. 1974. Body fat assessed from total body density and its

estimation from skinfold thickness: measurements on 481 men and women aged

from 16 to 72 years. Br J Nutr 32 (1):77-97.

455

Flegal, K. M., J. A. Shepherd, A. C. Looker, B. I. Graubard, L. G. Borrud, C. L. Ogden,

T. B. Harris, J. E. Everhart, and N. Schenker. 2009. Comparisons of percentage

body fat, body mass index, waist circumference, and waist-stature ratio in adults.

Am J Clin Nutr 89 (2):500-8.

Frisancho A.R, ed. 1990. Anthropometric Standards for the Assessment of Growth and

Nutritional Status. Ann Arbor: The University of Michigan Press.

Harries, A. D., W. A. Nkhoma, P. J. Thompson, D. S. Nyangulu, and J. J. Wirima. 1988.

Nutritional status in Malawian patients with pulmonary tuberculosis and response

to chemotherapy. Eur J Clin Nutr 42 (5):445-50.

Heitmann, B. L., H. Erikson, B. M. Ellsinger, K. L. Mikkelsen, and B. Larsson. 2000.

Mortality associated with body fat, fat-free mass and body mass index among 60-

year-old swedish men-a 22-year follow-up. The study of men born in 1913. Int J

Obes Relat Metab Disord 24 (1):33-7.

James, W. P., G. C. Mascie-Taylor, N. G. Norgan, B. R. Bistrian, P. S. Shetty, and A.

Ferro-Luzzi. 1994. The value of arm circumference measurements in assessing

chronic energy deficiency in Third World adults. Eur J Clin Nutr 48 (12):883-94.

Kennedy, N., A. Ramsay, L. Uiso, J. Gutmann, F. I. Ngowi, and S. H. Gillespie. 1996.

Nutritional status and weight gain in patients with pulmonary tuberculosis in

Tanzania. Trans R Soc Trop Med Hyg 90 (2):162-6.

456

Kotler, D. P., S. Burastero, J. Wang, and R. N. Pierson, Jr. 1996. Prediction of body cell

mass, fat-free mass, and total body water with bioelectrical impedance analysis:

effects of race, sex, and disease. Am J Clin Nutr 64 (3 Suppl):489S-497S.

Kyle, U. G., I. Bosaeus, A. D. De Lorenzo, P. Deurenberg, M. Elia, J. M. Gomez, B. L.

Heitmann, L. Kent-Smith, J. C. Melchior, M. Pirlich, H. Scharfetter, A. M.

Schols, and C. Pichard. 2004. Bioelectrical impedance analysis--part I: review of

principles and methods. Clin Nutr 23 (5):1226-43.

Kyle, U. G., I. Bosaeus, A. D. De Lorenzo, P. Deurenberg, M. Elia, J. Manuel Gomez, B.

Lilienthal Heitmann, L. Kent-Smith, J. C. Melchior, M. Pirlich, H. Scharfetter, M.

W. J. Schols A, and C. Pichard. 2004. Bioelectrical impedance analysis-part II:

utilization in clinical practice. Clin Nutr 23 (6):1430-53.

Kyle, U. G., L. Genton, and C. Pichard. 2002. Body composition: what's new? Curr Opin

Clin Nutr Metab Care 5 (4):427-33.

Kyle, U. G., A. Piccoli, and C. Pichard. 2003. Body composition measurements:

interpretation finally made easy for clinical use. Curr Opin Clin Nutr Metab Care

6 (4):387-93.

Lawn, S. D., and G. Churchyard. 2009. Epidemiology of HIV-associated tuberculosis.

Curr Opin HIV AIDS 4 (4):325-33.

Lean, M. E., T. S. Han, and P. Deurenberg. 1996. Predicting body composition by

densitometry from simple anthropometric measurements. Am J Clin Nutr 63

(1):4-14.

457

Lohman T.G, Roche A.F, and Martorell R, eds. 1988. Anthropometric Standardization

Reference Manual, Human Kinetics Books. Illinois: A Division of Human

Kinetics, Inc.

Lucas, S. B., K. M. De Cock, A. Hounnou, C. Peacock, M. Diomande, M. Honde, A.

Beaumel, L. Kestens, and A. Kadio. 1994. Contribution of tuberculosis to slim

disease in Africa. Bmj 308 (6943):1531-3.

Macallan, D. C. 1999. Malnutrition in tuberculosis. Diagn Microbiol Infect Dis 34

(2):153-7.

Mostert, R., A. Goris, C. Weling-Scheepers, E. F. Wouters, and A. M. Schols. 2000.

Tissue depletion and health related quality of life in patients with chronic

obstructive pulmonary disease. Respir Med 94 (9):859-67.

Mupere, E., S. Zalwango, A. Chiunda, A. Okwera, R. Mugerwa, and C. Whalen. 2010.

Body composition among HIV-seropositive and HIV-seronegative adult patients

with pulmonary tuberculosis in Uganda. Ann Epidemiol 20 (3):210-6.

Niyongabo, T., D. Henzel, M. Idi, S. Nimubona, E. Gikoro, J. C. Melchior, S. Matheron,

G. Kamanfu, B. Samb, B. Messing, J. Begue, P. Aubry, and B. Larouze. 1999.

Tuberculosis, human immunodeficiency virus infection, and malnutrition in

Burundi. Nutrition 15 (4):289-93.

Norgan, N. G. 2005. Laboratory and field measurements of body composition. Public

Health Nutr 8 (7A):1108-22.

458

Oosthuizen, G. M., G. Joubert, W. F. Mollentze, and E. Rosslee. 1997. Body fat

estimation in black South Africans: a pilot study. Cent Afr J Med 43 (5):126-31.

Paton, N. I., and Y. M. Ng. 2006. Body composition studies in patients with wasting

associated with tuberculosis. Nutrition 22 (3):245-51.

Powell-Tuck, J., and E. M. Hennessy. 2003. A comparison of mid upper arm

circumference, body mass index and weight loss as indices of undernutrition in

acutely hospitalized patients. Clin Nutr 22 (3):307-12.

Schutz, Y., U. U. Kyle, and C. Pichard. 2002. Fat-free mass index and fat mass index

percentiles in Caucasians aged 18-98 y. Int J Obes Relat Metab Disord 26

(7):953-60.

Schwenk A, and Macallan D.C. 2000. Tuberculosis, malnutrition and wasting. Curr Opin

Clin Nutr Metab Care 3:285 - 91.

Shah, S., C. Whalen, D. P. Kotler, H. Mayanja, A. Namale, G. Melikian, R. Mugerwa,

and R. D. Semba. 2001. Severity of human immunodeficiency virus infection is

associated with decreased phase angle, fat mass and body cell mass in adults with

pulmonary tuberculosis infection in Uganda. J Nutr 131 (11):2843-7.

Siri W.E. 1961. Body composition from fluid spaces and density: analyses of methods.

Washington, DC: National Academy of Sciences.

Tang, A. M., J. Forrester, D. Spiegelman, T. A. Knox, E. Tchetgen, and S. L. Gorbach.

2002. Weight loss and survival in HIV-positive patients in the era of highly active

antiretroviral therapy. J Acquir Immune Defic Syndr 31 (2):230-6.

459

Taylor, R. W., D. Keil, E. J. Gold, S. M. Williams, and A. Goulding. 1998. Body mass

index, waist girth, and waist-to-hip ratio as indexes of total and regional adiposity

in women: evaluation using receiver operating characteristic curves. Am J Clin

Nutr 67 (1):44-9.

van Lettow, M., W. W. Fawzi, and R. D. Semba. 2003. Triple trouble: the role of

malnutrition in tuberculosis and human immunodeficiency virus co-infection.

Nutr Rev 61 (3):81-90.

Van Lettow, M., J. J. Kumwenda, A. D. Harries, C. C. Whalen, T. E. Taha, N.

Kumwenda, C. Kang'ombe, and R. D. Semba. 2004. Malnutrition and the severity

of lung disease in adults with pulmonary tuberculosis in Malawi. Int J Tuberc

Lung Dis 8 (2):211-7.

VanItallie, T. B., M. U. Yang, S. B. Heymsfield, R. C. Funk, and R. A. Boileau. 1990.

Height-normalized indices of the body's fat-free mass and fat mass: potentially

useful indicators of nutritional status. Am J Clin Nutr 52 (6):953-9.

Villamor, E., E. Saathoff, F. Mugusi, R. J. Bosch, W. Urassa, and W. W. Fawzi. 2006.

Wasting and body composition of adults with pulmonary tuberculosis in relation

to HIV-1 coinfection, socioeconomic status, and severity of tuberculosis. Eur J

Clin Nutr 60 (2):163-71.

Wagner, G. J., S. J. Ferrando, and J. G. Rabkin. 2000. Psychological and physical health

correlates of body cell mass depletion among HIV+ men. J Psychosom Res 49

(1):55-7.

460

Zachariah, R., M. P. Spielmann, A. D. Harries, and F. M. Salaniponi. 2002. Moderate to

severe malnutrition in patients with tuberculosis is a risk factor associated with

early death. Trans R Soc Trop Med Hyg 96 (3):291-4.

461

CHAPTER FIVE

Altman D.G, ed. 1997. Practical Statistics for Medical Research. London: Chapman &

Hall.

Atkinson, S. A., and W. E. Ward. 2001. Clinical nutrition: 2. The role of nutrition in the

prevention and treatment of adult osteoporosis. CMAJ 165 (11):1511-4.

Biro, G., K. F. Hulshof, L. Ovesen, and J. A. Amorim Cruz. 2002. Selection of

methodology to assess food intake. Eur J Clin Nutr 56 Suppl 2:S25-32.

Cannon G. 2001. Diet-related chronic diseases. Focus 5 (brief 8 of 11):1 - 2.

Carloni, A. S. 1981. Sex disparities in the distribution of food within rural households.

Food Nutr (Roma) 7 (1):3-12. de Hartog A.P. 1972. Unequal distribution of food within the household: A somewhat

neglected aspect of food behavior. FAO Nutritional Newletter 10; 8 - 17.

Dey J. 1981. Gambian women: Unequal partners in rice development projects? In African

Women in the Development Process, edited by Nelson N. London: Frank Cass.

Dietary Reference Intake for Energy, Carbohydrate, Fiber, Fat, Fatty Acids, Cholesterol,

Protein, and Amino Acids (Macronutrients). A report of the Panel on

Macronutrients, Subcommittees on Upper Reference Levels of Nutrients and

Interpretation and Uses of Dietary Reference Intakes, and the Standing

462

Committee on the Scientific Evaluation of Dietary Reference Intakes. 2005.

Washington, DC: National Academies Press. www.nap.edu.

Food and Agriculture Organization (FAO)/World Health Organization (WHO). Human

Vitamin and Mineral Requirements. Report of a Joint FAO/WHO Expert

Consultation. 2002. Rome.

Food and Nutrition Board, National Research Council, National Academy of Sciences.

Recommended Dietary Allowances. 1989. 10th ed. Washington, DC: National

Academic Press.

Ghafoorunissa K.K. 1994. Diet and Heart Disease. Hyderabad: National Institute of

Nutrition.

Gopalan C, Rama Sastri BV, and Basasubramanian S.C. 1996. Nutritive Value of Indian

Foods. Hyderabad: National Institute of Nutrition, Indian Council of Medical

Research.

Habicht, J. P., L. D. Meyers, and C. Brownie. 1982. Indicators for identifying and

counting the improperly nourished. Am J Clin Nutr 35 (5 Suppl):1241-54.

Hatloy, A., L. E. Torheim, and A. Oshaug. 1998. Food variety--a good indicator of

nutritional adequacy of the diet? A case study from an urban area in Mali, West

Africa. Eur J Clin Nutr 52 (12):891-8.

Holmboe-Ottesen G, and Wandel M. 1991. Men's contribution to the food and nutritional

situation in the Tanzanian household. Ecology of Food and Nutrition 26:83 - 96.

463

Hopewell P.C. 1994. Overview of clinical tuberculosis. In Tuberculosis: Pathogenesis,

Protection and Control, edited by Bloom B.R. Washington D.C: ASM Press.

Lado, C. 1992. Female labour participation in agricultural production and the

implications for nutrition and health in rural Africa. Soc Sci Med 34 (7):789-807.

Madden, J. P., S. J. Goodman, and H. A. Guthrie. 1976. Validity of the 24-hr. recall.

Analysis of data obtained from elderly subjects. J Am Diet Assoc 68 (2):143-7.

O'Laughlin B. 1974. Mediation of contradiction: Why Mbum women do not eat chicken.

In Women, Culture and Society, edited by Rosaldo M.Z. and Lamphere L.

Stanford, CA: Stanford University Press.

Oldewage-Theron, W. H., and R. Kruger. 2008. Food variety and dietary diversity as

indicators of the dietary adequacy and health status of an elderly population in

Sharpeville, South Africa. J Nutr Elder 27 (1-2):101-33.

Paton, N. I., Y. K. Chua, A. Earnest, and C. B. Chee. 2004. Randomized controlled trial

of nutritional supplementation in patients with newly diagnosed tuberculosis and

wasting. Am J Clin Nutr 80 (2):460-5.

Rosenberg E.M. 1980. Demographic effects of sex-differential nutrition. In Nutritional

Anthropology: Contemporary Approaches to Diet and Culture, edited by Jerome

J.W., Kandel R.F. and Pelto G.H. Pleasantville, NY: Redgrave Publishing.

Ruel, M. T. 2003. Operationalizing dietary diversity: a review of measurement issues and

research priorities. J Nutr 133 (11 Suppl 2):3911S-3926S.

464

Sita-Lumbsden A, Lapthorn G, Swaminathan R, and Milburn H.J. 2007. Reactivation of

tuberculosis and vitamin D deficiency: the contribution of diet and exposure to

sunlight. Thorax 62 (11):1003 - 1007.

Steyn, N. P., J. H. Nel, G. Nantel, G. Kennedy, and D. Labadarios. 2006. Food variety

and dietary diversity scores in children: are they good indicators of dietary

adequacy? Public Health Nutr 9 (5):644-50.

Thurnham, D. I. 2004. An overview of interactions between micronutrients and of

micronutrients with drugs, genes and immune mechanisms. Nutr Res Rev 17

(2):211-40.

Tontisirin, K., G. Nantel, and L. Bhattacharjee. 2002. Food-based strategies to meet the

challenges of micronutrient malnutrition in the developing world. Proc Nutr Soc

61 (2):243-50.

van Lettow, M., W. W. Fawzi, and R. D. Semba. 2003. Triple trouble: the role of

malnutrition in tuberculosis and human immunodeficiency virus co-infection.

Nutr Rev 61 (3):81-90.

Vucic, V., M. Glibetic, R. Novakovic, J. Ngo, D. Ristic-Medic, J. Tepsic, M. Ranic, L.

Serra-Majem, and M. Gurinovic. 2009. Dietary assessment methods used for low-

income populations in food consumption surveys: a literature review. Br J Nutr

101 Suppl 2:S95-101.

WHO. 2002. World health report 2002 - reducing risks, promoting healthy life. Geneva:

World Health Organization.

465

CHAPTER SIX

Arciero, P. J., M. I. Goran, and E. T. Poehlman. 1993. Resting metabolic rate is lower in

women than in men. J Appl Physiol 75 (6):2514-20.

Baumgartner, R. N., K. M. Koehler, D. Gallagher, L. Romero, S. B. Heymsfield, R. R.

Ross, P. J. Garry, and R. D. Lindeman. 1998. Epidemiology of sarcopenia among

the elderly in New Mexico. Am J Epidemiol 147 (8):755-63.

Corbett, E. L., C. J. Watt, N. Walker, D. Maher, B. G. Williams, M. C. Raviglione, and

C. Dye. 2003. The growing burden of tuberculosis: global trends and interactions

with the HIV epidemic. Arch Intern Med 163 (9):1009-21.

Dales, L. G., and H. K. Ury. 1978. An improper use of statistical significance testing in

studying covariables. Int J Epidemiol 7 (4):373-5.

Eckel, L. A. 2004. Estradiol: a rhythmic, inhibitory, indirect control of meal size. Physiol

Behav 82 (1):35-41.

Fields-Gardner, C. 1995. A review of mechanisms of wasting in HIV disease. Nutr Clin

Pract 10 (5):167-76.

Garrow, J. S., and J. Webster. 1985. Quetelet's index (W/H2) as a measure of fatness. Int

J Obes 9 (2):147-53.

Geary, N. 2001. Estradiol, CCK and satiation. Peptides 22 (8):1251-63.

Geary, N. 2001. Sex differences in disease anorexia. Nutrition 17 (6):499-507.

466

Kennedy, N., A. Ramsay, L. Uiso, J. Gutmann, F. I. Ngowi, and S. H. Gillespie. 1996.

Nutritional status and weight gain in patients with pulmonary tuberculosis in

Tanzania. Trans R Soc Trop Med Hyg 90 (2):162-6.

Kotler, D. P. 2000. Nutritional alterations associated with HIV infection. J Acquir

Immune Defic Syndr 25 Suppl 1:S81-7.

Kotler, D. P., S. Burastero, J. Wang, and R. N. Pierson, Jr. 1996. Prediction of body cell

mass, fat-free mass, and total body water with bioelectrical impedance analysis:

effects of race, sex, and disease. Am J Clin Nutr 64 (3 Suppl):489S-497S.

Kotler, D. P., and C. Grunfeld. 1995. Pathophysiology and treatment of the AIDS wasting

syndrome. AIDS Clin Rev:229-75.

Kyle, U. G., I. Bosaeus, A. D. De Lorenzo, P. Deurenberg, M. Elia, J. Manuel Gomez, B.

Lilienthal Heitmann, L. Kent-Smith, J. C. Melchior, M. Pirlich, H. Scharfetter, M.

W. J. Schols A, and C. Pichard. 2004. Bioelectrical impedance analysis-part II:

utilization in clinical practice. Clin Nutr 23 (6):1430-53.

Kyle, U. G., L. Genton, and C. Pichard. 2002. Body composition: what's new? Curr Opin

Clin Nutr Metab Care 5 (4):427-33.

Kyle, U. G., A. Piccoli, and C. Pichard. 2003. Body composition measurements:

interpretation finally made easy for clinical use. Curr Opin Clin Nutr Metab Care

6 (4):387-93.

467

Kyle, U. G., Y. Schutz, Y. M. Dupertuis, and C. Pichard. 2003. Body composition

interpretation. Contributions of the fat-free mass index and the body fat mass

index. Nutrition 19 (7-8):597-604.

Lawn, S. D., and G. Churchyard. 2009. Epidemiology of HIV-associated tuberculosis.

Curr Opin HIV AIDS 4 (4):325-33.

Lennie, T. A. 2004. Sex differences in severity of inflammation-induced anorexia and

weight loss. Biol Res Nurs 5 (4):255-64.

Lucas, S. B., K. M. De Cock, A. Hounnou, C. Peacock, M. Diomande, M. Honde, A.

Beaumel, L. Kestens, and A. Kadio. 1994. Contribution of tuberculosis to slim

disease in Africa. Bmj 308 (6943):1531-3.

Macallan, D. C. 1999. Malnutrition in tuberculosis. Diagn Microbiol Infect Dis 34

(2):153-7.

Macallan, D. C., M. A. McNurlan, A. V. Kurpad, G. de Souza, P. S. Shetty, A. G. Calder,

and G. E. Griffin. 1998. Whole body protein metabolism in human pulmonary

tuberculosis and undernutrition: evidence for anabolic block in tuberculosis. Clin

Sci (Lond) 94 (3):321-31.

Mupere, E., S. Zalwango, A. Chiunda, A. Okwera, R. Mugerwa, and C. Whalen. 2010.

Body composition among HIV-seropositive and HIV-seronegative adult patients

with pulmonary tuberculosis in Uganda. Ann Epidemiol 20 (3):210-6.

468

Neter J., Wasserman W., and Kutner M.H. 1990. Applied linear statistical models:

regression, analysis of variance, and experimental designs. 3rd ed. Homewood,

IL: Richard D Irwin Inc.

Niyongabo, T., D. Henzel, M. Idi, S. Nimubona, E. Gikoro, J. C. Melchior, S. Matheron,

G. Kamanfu, B. Samb, B. Messing, J. Begue, P. Aubry, and B. Larouze. 1999.

Tuberculosis, human immunodeficiency virus infection, and malnutrition in

Burundi. Nutrition 15 (4):289-93.

Paton, N. I., L. R. Castello-Branco, G. Jennings, M. B. Ortigao-de-Sampaio, M. Elia, S.

Costa, and G. E. Griffin. 1999. Impact of tuberculosis on the body composition of

HIV-infected men in Brazil. J Acquir Immune Defic Syndr Hum Retrovirol 20

(3):265-71.

Paton, N. I., and Y. M. Ng. 2006. Body composition studies in patients with wasting

associated with tuberculosis. Nutrition 22 (3):245-51.

Paton, N. I., Y. M. Ng, C. B. Chee, C. Persaud, and A. A. Jackson. 2003. Effects of

tuberculosis and HIV infection on whole-body protein metabolism during feeding,

measured by the [15N]glycine method. Am J Clin Nutr 78 (2):319-25.

Rammohan, M., K. Kalantar-Zadeh, A. Liang, and C. Ghossein. 2005. Megestrol acetate

in a moderate dose for the treatment of malnutrition-inflammation complex in

maintenance dialysis patients. J Ren Nutr 15 (3):345-55.

Rubin, S. A. 1995. Tuberculosis. Captain of all these men of death. Radiol Clin North Am

33 (4):619-39.

469

Shah, S., C. Whalen, D. P. Kotler, H. Mayanja, A. Namale, G. Melikian, R. Mugerwa,

and R. D. Semba. 2001. Severity of human immunodeficiency virus infection is

associated with decreased phase angle, fat mass and body cell mass in adults with

pulmonary tuberculosis infection in Uganda. J Nutr 131 (11):2843-7.

Tarnopolsky, M. A. 2000. Gender differences in metabolism; nutrition and supplements.

J Sci Med Sport 3 (3):287-98.

Tarnopolsky, M. A. 2000. Gender differences in substrate metabolism during endurance exercise. Can J Appl Physiol 25 (4):312-27.

Tarnopolsky, M. A. 2008. Sex differences in exercise metabolism and the role of 17-beta

estradiol. Med Sci Sports Exerc 40 (4):648-54.

Tarnopolsky, M. A., and B. C. Ruby. 2001. Sex differences in carbohydrate metabolism.

Curr Opin Clin Nutr Metab Care 4 (6):521-6.

Van Lettow, M., J. J. Kumwenda, A. D. Harries, C. C. Whalen, T. E. Taha, N.

Kumwenda, C. Kang'ombe, and R. D. Semba. 2004. Malnutrition and the severity

of lung disease in adults with pulmonary tuberculosis in Malawi. Int J Tuberc

Lung Dis 8 (2):211-7.

VanItallie, T. B., M. U. Yang, S. B. Heymsfield, R. C. Funk, and R. A. Boileau. 1990.

Height-normalized indices of the body's fat-free mass and fat mass: potentially

useful indicators of nutritional status. Am J Clin Nutr 52 (6):953-9.

Villamor, E., E. Saathoff, F. Mugusi, R. J. Bosch, W. Urassa, and W. W. Fawzi. 2006.

Wasting and body composition of adults with pulmonary tuberculosis in relation

470

to HIV-1 coinfection, socioeconomic status, and severity of tuberculosis. Eur J

Clin Nutr 60 (2):163-71.

Vittinghoff E., Shiboski S.C., Glidden D.V, and McCulloch C.E. 2005. Regreession

Methods in Biostatistics, Linear, Logistic, Survival, and Repeated Measure

Models. Edited by Gail M., Krickeberg K., Wong W., Samet J. and Tsiatis A.

New York, NY: Springer Science + Business Media.

World Health Organization. Global tuberculosis control. Epidemiology, strategy,

financing. WHO/HTM/TB/2009.411. 2009. Geneva: World Health Organization

471

CHAPTER SEVEN

Arciero, P. J., M. I. Goran, and E. T. Poehlman. 1993. Resting metabolic rate is lower in

women than in men. J Appl Physiol 75 (6):2514-20.

Baumgartner, R. N., K. M. Koehler, D. Gallagher, L. Romero, S. B. Heymsfield, R. R.

Ross, P. J. Garry, and R. D. Lindeman. 1998. Epidemiology of sarcopenia among

the elderly in New Mexico. Am J Epidemiol 147 (8):755-63.

Carloni, A. S. 1981. Sex disparities in the distribution of food within rural households.

Food Nutr (Roma) 7 (1):3-12. de Hartog A.P. 1972. Unequal distribution of food within the household: A somewhat

neglected aspect of food behavior. FAO Nutritional Newletter 10; 8 - 17.

Dey J. 1981. Gambian women: Unequal partners in rice development projects? In African

Women in the Development Process, edited by Nelson N. London: Frank Cass.

Duarte, E. C., A. L. Bierrenbach, J. Barbosa da Silva, Jr., P. L. Tauil, and E. de Fatima

Duarte. 2009. Factors associated with deaths among pulmonary tuberculosis

patients: a case-control study with secondary data. J Epidemiol Community

Health 63 (3):233-8.

Heitmann, B. L., H. Erikson, B. M. Ellsinger, K. L. Mikkelsen, and B. Larsson. 2000.

Mortality associated with body fat, fat-free mass and body mass index among 60-

year-old swedish men-a 22-year follow-up. The study of men born in 1913. Int J

Obes Relat Metab Disord 24 (1):33-7.

472

Holmboe-Ottesen G, and Wandel M. 1991. Men's contribution to the food and nutritional

situation in the Tanzanian household. Ecology of Food and Nutrition 26:83 - 96.

Khan, A., T. R. Sterling, R. Reves, A. Vernon, and C. R. Horsburgh. 2006. Lack of

weight gain and relapse risk in a large tuberculosis treatment trial. Am J Respir

Crit Care Med 174 (3):344-8.

Kotler, D. P., S. Burastero, J. Wang, and R. N. Pierson, Jr. 1996. Prediction of body cell

mass, fat-free mass, and total body water with bioelectrical impedance analysis:

effects of race, sex, and disease. Am J Clin Nutr 64 (3 Suppl):489S-497S.

Kyle, U. G., L. Genton, and C. Pichard. 2002. Body composition: what's new? Curr Opin

Clin Nutr Metab Care 5 (4):427-33.

Kyle, U. G., A. Piccoli, and C. Pichard. 2003. Body composition measurements:

interpretation finally made easy for clinical use. Curr Opin Clin Nutr Metab Care

6 (4):387-93.

Lado, C. 1992. Female labour participation in agricultural production and the

implications for nutrition and health in rural Africa. Soc Sci Med 34 (7):789-807.

Lawn, S. D., and G. Churchyard. 2009. Epidemiology of HIV-associated tuberculosis.

Curr Opin HIV AIDS 4 (4):325-33.

Lucas, S. B., K. M. De Cock, A. Hounnou, C. Peacock, M. Diomande, M. Honde, A.

Beaumel, L. Kestens, and A. Kadio. 1994. Contribution of tuberculosis to slim

disease in Africa. Bmj 308 (6943):1531-3.

473

Macallan, D. C. 1999. Malnutrition in tuberculosis. Diagn Microbiol Infect Dis 34

(2):153-7.

Muller, O., and M. Krawinkel. 2005. Malnutrition and health in developing countries.

CMAJ 173 (3):279-86.